Ideological Drift among Supreme Court Justices Who, When, and

Document Sample
Ideological Drift among Supreme Court Justices Who, When, and Powered By Docstoc
					                Ideological Drift among Supreme Court Justices:
                       Who, When, and How Important?∗

              Lee Epstein, Andrew D. Martin, Kevin M. Quinn & Jeffrey A. Segal†



                                                    Abstract

           When President George W. Bush declared that his Supreme Court nominee, Harriet E.
       Miers, was “not going to change, that 20 years from now she’ll be the same person with the
       same philosophy that she is today,” no one should have been shocked. To the contrary: The
       President was merely reiterating a claim dominant in public and scholarly discourse over the
       Supreme Court—that justices come to the Court with robust ideological outlooks and do not
       veer from them over the course of their tenure. Nonetheless, and despite the commonplace
       nature of the claim, it is not without its share of skeptics; indeed, some commentators now
       contend that ideological drift among Supreme Court justices is not just possible but likely.
           Using systematically developed data and sophisticated statistical tools, we address the ques-
       tion of whether justices remain committed to a particular doctrinal course over time. The
       results, as it turns out, could not be clearer: Contrary to the received wisdom, virtually all
       justices serving since 1937 has grown more liberal or conservative during their tenure on the
       Court.
           Finding that change is the rule, not the exception, we develop the implications of our findings
       for the justices’ appointment to the Court and the doctrine they develop once confirmed. We
       show, for example, that Presidents hoping to create a lasting legacy in the form of justices who
       share their ideology can be reasonably certain that their appointees will behave in line with
       contemporaneous expectations—at least during the justice’s first term in office. But because
       most justices fluctuate soon thereafter, Presidents emphasize ideology to the neglect of other
       considerations—such as the advancement of their political party’s electoral ambitions—at their
       own peril.
           We conclude with a discussion of the prospects for legal change among the justices of the
       Roberts Court. Here we consider two plausible scenarios, one in which the justices remain
       relatively true to their current doctrinal inclinations and another in which members drift. Either
       way, we find that legal change may be possible—a finding that defies contemporary expectations
       about the inertia of justices and, by implication, the Court in the absence of membership
       turnover.


   ∗
     Forthcoming, Northwestern University Law Review (2007).
   †
     Lee Epstein (http://epstein.law.northwestern.edu/) is the Beatrice Kuhn Professor of Law and Professor Po-
litical Science at Northwestern University; Andrew D. Martin is Professor of Law and Political Science at Washington
University in St. Louis; Kevin M. Quinn is Assistant Professor of Government at Harvard University and a Fellow at
the Center for Advanced Studies in the Behavioral Sciences; Jeffrey A. Segal is Distinguished Professor of Political
Science at Stony Brook University. We are grateful to Linda Greenhouse for motivating our analysis of ideological
change, as well as for her insights on the subject; to the National Science Foundation and Northwestern University
School of Law for supporting our work; and to Barry Friedman, John McGinnis, and Nancy Staudt for sharing their
thoughts on the contemporary Court. The project’s web site houses a full replication archive (http://epstein.law.
northwestern.edu/research/ideodrift.html).
                                                I    Introducton

   When the U.S. Supreme Court upheld the use of military commissions for enemy combatants
in Hamdan v. Rumsfeld,1 the decision fueled more than a national debate over the powers of
the President. It also generated commentary about the ideological composition of the Court.
Conservatives proclaimed that they were just one justice, just one vacancy, away from victory in
Hamdan 2 and a handful of other recent decisions that worked against their interests.3 Liberals
worried about just as much.4
   The commentary over Hamdan reflects a widely shared belief among journalists, politicians,
scholars, and even judges: alterations in the Court’s jurisprudence are unlikely in the absence of
membership change. That is because the justices themselves do not exhibit ideological change
over the course of their tenure.5 To paraphrase the old proverb, once a conservative, always a
conservative. Likewise for liberals.6
     Why the assumption of stable preferences is so deeply held is open to speculation. Some analysts
suggest it would defy logic to expect mature persons, with years of experience in the legal world,
to revisit their jurisprudential views. Would a John G. Roberts, Jr.—a justice who has studied,
litigated, or adjudicated court cases for over half his life—alter his ideological preferences? The
answer, according to Professor David A. Strauss, is that he would not:
   1
     126 S. Ct. 2749 (2006).
   2
     The vote in Hamdan was five-to-three. Because he served on the appellate court panel that had upheld the
commissions, Hamdan v. Rumsfeld, 415 F.3d 33 (D.C. Cir. 2005), Chief Justice Roberts recused himself. Had he
participated, many commentators assume he would have once again supported the administration. See, e.g., Cass
Sunstein, The Court’s Stunning Hamdan Decision, New Rep. Online, June 30, 2006, http://www.tnr.com/ (“The
current Court itself remains badly divided. We should emphasize that Hamdan was decided by a narrow margin of
5-3, and we should not neglect the fact that Chief Justice Roberts did not participate in the decision; the reason is
that he was part of the three-judge lower court, now reversed, which had ruled broadly in the President’s favor.”)
   3
     E.g., the five-to-four decisions in Kelo v. City of New London, 545 U.S. 469 (2005) (taking of property for
economic development does not violate the “public use” restriction of the Fifth Amendment’s Taking Clause); Grutter
v. Bollinger, 539 U.S. 306 (2003) (a law school’s use of race in admissions decisions does not violate the Fourteenth
Amendment’s Equal Protection Clause; Roper v. Simmons, 543 U.S. 551 (2005) (the Eighth Amendment prohibits
the imposition of the death penalty for crimes committed when the defendant was under the age of 18).
   4
     Commentary on Hamdan and the ideological composition of the Court appears on numerous blogs. See, e.g., the
Journal of Applied Episdemology, Today’s Hamdan Decision, June 30, 2006, http://appliedepistemology.com/node/96
(“The scary lesson that Hamdan teaches us is that the only thing currently standing between American democracy
and an executive branch autocracy is John Paul Stevens’ bath mat.”); the National Review Online, Five, Wrong on
Hamdan, June 30, 2006, http://article.nationalreview.com/ (“The Mystery Five [justices] have simply practiced once
again the utterly lawless willfulness that they have proclaimed to be their mission. And they undoubtedly know that
they will receive ample cover, in the form of fawning accolades, from legal academia and the liberal media.”).
   5
     We develop these points infra Part IIA; see also infra note 25. Suffice it to note here that the claim of ideological
consistency not only appears in commentary on the Court but undergirds many important theories of judicial deci-
sions, or at least tests of those theories. Consider “separation of powers” theories, which suggest that the Court takes
into account the preferences and likely actions of Congress when it interprets statutes. The typical assumption is that
the sincere preferences of Court do not change unless the center of the Court (the median) changes as a result of mem-
bership turnover. See, e.g., William N. Eskridge, Jr., Overriding Supreme Court Statutory Interpretation Decisions,
101 Yale L. J. 331 (1991); William N. Eskridge, Jr., Reneging on History? Playing the Court/Congress/President
Civil Rights Game, 79 Calif. L. Rev. 613 (1991); Pablo Spiller & Rafael Gely, Congressional Control of Judi-
cial Independence: The Determinants of U.S. Supreme Court Labor-Relations Decisions, 1949-1988, 23 RAND J.
Econ. 463 (1992) (all three detailing how the Court’s sincere or raw preferences move with membership changes
but explaining why the Court may not act on those preference). Likewise, some adherents of the attitudinal model
of judicial decisions, which holds that justices vote on the basis of their ideology, describe attitudes as “relatively
enduring.” See David W. Rohde & Harold J. Spaeth, Supreme Court Decision Making (1976), 72.
   6
     The proverb is “Once a thief, always a thief”.



                                                           1
       As Americans try to figure out what Judge John G. Roberts Jr. will be like as a U.S.
       Supreme Court justice, one idea seems to [be] that whatever Judge Roberts is now, once
       he is on the court he might develop into something different. In particular, the thinking
       goes, even if he is the intense conservative suggested by his Reagan-era memoranda, he
       may become more moderate as a justice.
       Don’t believe it.7

    Shoring up intuitions about the implausibility of preference change is empirical support in the
form of a William H. Rehnquist on the right and a Thurgood Marshall on the left—justices who
never seemed to veer from their preferred ideological course. When President Richard Nixon ap-
pointed Rehnquist to the Court, virtually all observers of the day deemed the nominee a reliable
conservative.8 Likewise at the time of his appointment, the press declared Justice Marshall a prob-
able addition to the Court’s “liberal bloc.”9 That these initial ideological labels well characterized
the justices’ future behavior only serves to confirm Professor Strauss’s claim about the unlikelihood
of change. Or so the argument goes.
    And yet, despite the commonplace nature of the claim, it is not without its share of skeptics.
Whether pointing to anecdotes or more systematic evidence, several analysts now contend that
ideological drift is not just possible but likely.10 Exhibit A, they say, is Harry A. Blackmun. While
the justice himself maintained that it was the Court, not he, that moved—“I don’t believe I’m
any more liberal, as such, now than I was before,” Justice Blackmun once told a reporter11 —many
scholars disagree.12 To them, it is hard to believe that the same justice who dissented from the
Court’s 1972 decision to strike down existing death penalty statutes13 wrote, in 1994, “From this
day foward, I no longer shall tinker with the machinery of death.”14
     But is Justice Blackmun the rule or the rare exception? Do most justices remain committed
to a particular doctrinal course throughout their careers, as Strauss and others contend, or do the
skeptics have the better case? After reviewing the relevant commentary in Part II, we deploy state-
of-the-art methods to address these questions. The results, as it turns out, could not be clearer:
Contrary to the received wisdom, virtually every justice serving since the 1930s has moved to the
left or right or, in some cases, has switched directions several times.
    Finding that ideological drift is pervasive, in Part IV we develop the implications of our results
for two moments in the justices’ career cycle: the events surrounding their appointment to the
   7
        David A. Strauss, It’s Time to Deal with Reality: The Myth of the Unpredictable Supreme Court Justice Debunked,
Chi. Trib., Aug. 7, 2005.
      8
        See infra Part IIA.
      9
        Louis Hohlmeier, Thurgood Marshall Chosen for High Court; First Negro will Bolster Liberal Segment, Wall St.
J., June 14, 1967, at 3.
    10
        See Theodore W. Ruger, Justice Harry Blackmun and the Phenomenon of Judicial Preference Change, 70 Mo.
L. Rev. 1209, 1220 (2005) (a “small but emerging body of empirical literature suggests that preference change in
a phenomenon which affects many justices over the course of their careers.”). See also infra Part II for a review of
studies suggesting that justices change over time.
    11
        Quoted in John A. Jenkins, A Candid Talk with Justice Blackmun, NY Times, Feb. 20, 1983, 26.
    12
        See, e.g., Note, The Changing Social Vision of Justice Harry A. Blackmun: The Conscientious Conscience,
96 Harv. L. Rev. 717, 717 (1983) (“Harry A. Blackmun has undergone a remarkable transformation”); Linda
Greenhouse, Becoming Justice Blackmun: Harry Blackmun’s Supreme Court Journey 235 (2005) (pre-
senting data documenting Blackmun’s movement.); Ruger supra note 10, at 1212 (“diverse evidence suggests that
. . . Blackmun significantly changed.”).
    13
        Furman v. Georgia, 404 U.S. 238 (192)
    14
        Callins v. Collins, 510 U.S. 1141, 1145 (1994) (Blackmun, J., dissenting).


                                                          2
Court and the doctrine they develop once confirmed. As to the first, we show that Presidents
hoping to create a lasting legacy in the form of justices who share their ideology can be reasonably
certain that their appointees will behave in line with expectations—at least during the justice’s
first term in office. But, even before hitting the first-decade mark, most justices fluctuate, leading
to a degradation of the relationship between their preferences and their votes. The implication
is clear: Contrary to the claims of prominent scholars, the President and his supporters in the
Senate cannot guarantee the “entrenchment” of their ideology on the Court in the long or even
medium term.15 As a result, Presidents may be best off placing comparatively greater emphasis on
advancing the interests of their political party—rather than their own ideological interests—through
the appointment of justices designed to appease particular constituencies.
    As for the development of doctrine, contrary to the prevailing wisdom, we find that ideological
movement can manifest in important legal change. To provide but one example, had Justice Sandra
Day O’Connor’s initial preferences remained stable, odds are that she would not have provided the
fifth vote to uphold Michigan Law School’s affirmative action program in the 2003 case, Grutter
v. Bollinger.16 The implications of this finding are many, not the least of which is that attorneys’
expectations about success (or failure) with particular justices may rest on shakier ground than
they suspect.
    We conclude, in Part V with a discussion of the prospects for legal change among the justices
of the Roberts Court. Here we consider two plausible scenarios, one in which the current justices
remain relatively true to their current doctrinal inclinations and another in which members move.
Either way, we find that legal change (or, in some instances, surprising stability) may be possible—a
finding that defies contemporary expectations about the inertia of justices and, by implication, the
Court in the absence of membership turnover.


   II    Change on the Court: Conventional Views, Challenges, and Implications

   When Professor Strauss implies that John G. Roberts, Jr. will not and, in fact, most justices
do not change their ideological outlook with time, he expresses the conventional view of judging.
Indeed, even before their confirmation, journalists, scholars, and naturally enough policy makers
place justices into one ideological box or another, and assume that they will stay put over the
  15
    Jack M. Balkin & Sanford Levinson, Understanding the Constitutional Revolution, 87 Va. L. Rev. 1045, 1067
(2001) explain their “theory of partisan entrenchment” in the following terms:
        When a party wins the White House, it can stock the federal judiciary with members of its own party,
        assuming a relatively acquiescent Senate. They will serve for long periods of time because judges enjoy
        life tenure. On average, Supreme Court Justices serve about eighteen years. In this sense, judges
        and Justices resemble Senators who are appointed for 18-year terms by their parties and never have
        to face election. They are temporally extended representatives of particular parties, and hence, of
        popular understandings about public policy and the Constitution. The temporal extension of partisan
        representation is what we mean by partisan entrenchment.
As other scholars have recognized, a finding of widespread preference change would present a serious challenge
to theories of partisan or ideological entrenchment. See, e.g., Lee Epstein & Jeffrey A. Segal, Advice and
Consent: The Politics of Judicial Appointment (2005), 141 (“Whether or not packing the courts is a laudable
goal, a variety of factors can conspire against Presidents seeking to achieve it,” including “changing attitudes.”);
Ruger, supra note 10, at 1211 (“The possibility that judicial preferences might vary significantly over time compels
reconsideration of . . . entrenchment theory.”)
  16
     539 U.S. 306 (2003).




                                                          3
course of their tenure.17
    To see the point we need only consider the most recent appointee, Samuel A. Alito, Jr. From
the day President Bush announced the nomination, newspapers as ideological disparate as Wall
Street Journal and the New York Times deemed Alito a “right-of-center” nominee. “With yes-
terday’s nomination of Sam Alito to the Supreme Court,” wrote the Journal’s editors, “President
Bush reached into his John Roberts’ playbook to name a judicial conservative with impeccable
credentials.”18 The liberal Times agreed:

       The [President’s] solution to almost every problem seems to be either to rely on a
       close personal associate or to pander to his right wing. When the first tactic failed to
       work with the Harriet Miers nomination, Mr. Bush resorted to the second. The Alito
       nomination has thrilled social conservatives, who regard the judge to be a surefire vote
       against abortion rights.”19

After the Senate’s hearings, the editors of both papers became even more secure in their predictions.
“What we’re confident Judge Alito won’t do,” proclaimed the Journal, “is join the Court’s liberal
wing on cases such as Lawrence [v. Texas], and intrude willy-nilly into social matters best left to
legislatures to solve.”20 The Times even advocated a filibuster because of “Judge Alito’s refusal to
even pretend to sound like a moderate.”21
    Clearly the assumption that Alito was a conservative and would remain a conservative dom-
inated contemporary discourse, as it has over so many recent nominations. Nonetheless, at least
some commentators question the assumption of ideological stability.22 Both doctrinal and empirical
analyses, they assert, support the view that justices can and do change over the course of their
tenure. They even contend that ideological movement is possible for those justices, such as Alito,
who appear solidly in one ideological camp or the other.23
    In what follows we briefly consider the conventional assumption about the lack of ideological
movement and challenges to it. We end with a consideration of why this debate is worthwhile to
resolve.
   17
      See, e.g., Richard G. Wilkins, et al. Supreme Court Voting Behavior: 2003 Term, 32 Hastings Const. L.Q. 769,
776 (“both the media and academicians are fond of attaching ideological labels to the Court and its personnel.”);
Ruger, supra note 10, at 1209-1210 (“We are fond of putting our [justices] into neat adjectival boxes. . . . These ty-
pologies often reflect perceived attitudinal or ideological preferences: some justices are called ‘liberal’ or ‘conservative’
or ‘moderate’ . . . . But efforts to describe and classify the Justices . . . often rely . . . on the idea that once a Justice is
properly pegged, his or ideology . . . is not expected to evolve much.”); Robert E. Riggs, When Every Vote Counts:
5-4 Decisions in the United States Supreme Court, 1900-90, 21 Hofstra L. Rev. 667, 701 (1993) (“assigning
ideological labels” is appropriate because “during most Terms, most Justices voted consistently with their labels.”)
   18
      A Fight Worth Having, Wall St. J., Nov. 1, 2005, at A16.
   19
      Another Lost Opportunity, NY Times, November 1, 2005, at A26.
   20
      Hot Topic: Stare Indecisis, Wall St. J., January 14, 2006, at A8.
   21
      The editorial continued: “A filibuster is a radical tool. It’s easy to see why Democrats are frightened of it. But
from our perspective, there are some things far more frightening. One of them is Samuel Alito on the Supreme
Court.” Senators in Need of a Spine, NY Times, Jan. 26, 2006, 22A.
   22
      CITE needed here. See infra.
   23
      See infra, Part IIB.




                                                               4
    A    The Conventional View

    As even our brief discussion thus far suggests, no one should have been shocked when President
George W. Bush declared that his Supreme Court nominee, Harriet E. Miers, was “not going to
change, that 20 years from now she’ll be the same person with the same philosophy that she is
today.”24 To the contrary: The President was merely reiterating an assumption dominant in public
and scholarly discourse on the Supreme Court—what we call the assumption of stability, or the
idea that justices come to the Court with robust ideological outlooks and do not veer from them
over the course of their tenure.25
     The genesis of this view seems to lie both in intuition and empirical observation. Intuitively,
it seems implausible to believe that justices would have pause to rethink their presumably well-
entrenched beliefs over matters jurisprudential. Consider Ruth Bader Ginsburg. As a former law
professor, she presumably held strong views about the areas of law in which she taught, wrote, and
litigated; it is the odd law professor who does not, and Ginsburg appears to be no exception.26 As
a U.S. Court of Appeals judge, she likely held or developed preferences over the wide array of legal
matters she adjudicated; it is the odd judge who does not.27 Moving up to the Supreme Court,
under most theories of judging, would give her even more freedom to act on those preferences, and
act on them term after term.28
    Justice Ginsburg, of course, is not alone. In looking at the thirty-six justices who have served
since 1937,29 twenty were law professors or judges at the time of their nomination30 —including
each and every member of the current Court. On average, the justices serving in the 2006 term
sat as federal appellate judges for seven years. The three former law professors, Justices Scalia,
Breyer, Ginsburg, worked in the academy for a combined total of thirty-seven years.31
   24
      Press conference, October 4, 2005. Transcript available at:
http://www.whitehouse.gov/news/releases/2005/10/20051004-1.html.
   25
      Ruger, supra note 10, at 1218, deems the assumption of preference stability “near hegemonic.” With the scattered
exceptions we review in infra Part IIB, we wholeheartedly concur with Ruger’s sentiment. The assumption lies at
the core of many theories of judicial decision making, or at least the tests of those theories. For examples, see supra
note 5, as well as Ruger, supra note 10, at 1217-1218. It has been repeated in many scholarly studies of the Court,
as well as in more informal commentary. See, e.g., Strauss, supra note 7; Lawrence Baum, Measuring Policy Change
in the U.S. Supreme Court, 82 Am. J. Pol. Sci. 905, 907 (develops a method for assessing policy change based on
the assumption that justices’ preferences “remain constant throughout [their] career”); Glendon Schubert, The
Judicial Mind Revisited (1974), 159, (presents data showing a high-level of stability in voting for justices serving
from 1946-1968). .
   26
      Actually, prior to her service on the U.S. Court of Appeals for the District of Columbia, Justice Ginsburg was
a prominent and unabashed supporter of women’s rights and a pro-choice advocate. Among her many writings on
these subjects are The Equal Rights Amendment is the Way, 1 Harv. Womens L. 19 (1978); Sexual Equality Under
the Fourteenth and Equal Rights Amendments, 1979 Wash. U. L. Q. 161 (1979).
   27
      For analyses of the effect of ideology or partisanship on judging at the federal appellate level, see Cass R.
Sunstein, et al. Are Judges Political: An Empirical Analysis of the Federal Judiciary (2006); Frank
B. Cross & Emerson H. Tiller, Judicial Partisanship and Obedience to Legal Doctrine: Whistleblowing on the Federal
Courts of Appeals, 107 Yale L. J. 2155 (1998).
   28
      E.g., on the attitudinal model of judging, justices vote on the basis of their sincerely held ideological attitudes
toward cases before them. Freeing justices from considerations other than ideology, according to attitudinalists, is the
lack of electoral accountability and ambition for higher office, the control they enjoy over their agenda, and the dearth
of judicial superiors. See Jeffrey A. Segal & Harold J. Spaeth, The Supreme Court and the Attitudinal
Model Revisited (2002).
   29
      To derive the figure of thirty-six we count Chief Justice Rehnquist only once.
   30
      Data in this paragraph are derived from Lee Epstein, et al., The Supreme Court Compendium (2007),
Table 4-12.
   31
      Justice Breyer, at Harvard from 1967-70; Justice Ginsburg at Rutgers, 1963-72 and Columbia 1972-80; and



                                                           5
    Lending weight to intuitions about the entrenchment of ideology, and thus the implausibility
of change, comes a wealth of behavioral data. Particular impressive is the extent to which initial
impressions of the ideology of the justices, as nominees, correlate with their subsequent voting
on the Court. At the time of his appointment, as we noted, journalists deemed Samuel Alito
a conservative.32 Three decades earlier, newspaper editors wrote much the same of the Richard
Nixon nominee, William H. Rehnquist. “Mr. Rehnquist,” according to the New York Times was
“a Goldwater conservative [with] a brilliant professional background but a questionable record on
civil liberties.”33 And twenty years before Rehnquist, the press pigeon-holed William J. Brennan,
Jr. as a liberal.34
    The newspaper editors were hardly in error. Over the course of his thirty-five years of service
Chief Justice Rehnquist supported defendants in only two out of every ten criminal cases, and civil
rights plaintiffs in but 27 percent of the 694 discrimination suits in which he participated.35 Those
figures for Brennan were nearly the reverse. In only 20 percent of the cases did he vote against
defendants or civil rights plaintiffs. As for Justice Alito, his voting in the 2005 term places him
closer to a Rehnquist than a Brennan, just as the editors predicted.36
    It is one thing, of course, for the press to forecast accurately the behavior of a few seemingly
extreme ideologues, the Rehnquists and Brennans, and quite another to predict the voting of
the balance of nominees—some of whom had said or written little prior to their appointment.
Nonetheless, the newspaper editors generally meet that more rigorous standard, as Figure 1 shows.
There we draw a comparison between the editors’ initial branding of the Supreme Court nominees
(as analyzed and summarized by Jeffrey A. Segal, et al.) and the votes they, as justices, later
cast.37 Specifically, on the horizontal axis we display the editors’ ideological assessments, ranging
from very liberal to very conservative.38 Note that nominees deemed conservative by the journalists
appear toward the right of the figure (e.g., Chief Justice Rehnquist); liberals are toward the left
(e.g., Justice Brennan). On the vertical axis we show the percentage of conservative votes cast by
the justice over the course of his or her career.39 Justices who cast a high percentage of conservative
Justice Scalia at Virginia, 1967-74 and Chicago, 1977-82. We include here only full-time service. Justice Alito
served as an Adjunct Professor at Seton Hall from 1999-04; Justice Kennedy lectured at the University of the Pacific
between 1965-68; and Chief Justice Roberts was an adjunct at Georgetown in 2005. Justice Stevens was a lecturer
at Northwestern, 1950-54 and at Chicago 1954-58.
  32
     See supra notes 18, 19, 20, and 21.
  33
     The Court Nominations, NY Times, Oct. 22, 1971, at 38.
  34
     See infra Figure 1.
  35
     We computed the figures in this paragraph from the Harold J. Spaeth’s U.S. Supreme Court Database, with
analu=0 and dec type=1, 6, or 7, available at http://www.as.uky.edu/polisci/ulmerproject/sctdata.htm.
  36
     According to Spaeth’s database, supra note 35, in the 2005 term Alito supported criminal defendants in 16.7
percent of 12 cases in which he participated; he supported civil rights plaintiffs in 3 of 5 cases. See also infra Figure
19.
  37
     Jeffrey A. Segal, et al. create their editors’ ideology scores by content analyzing the editorials in four newspapers—
two with liberal leanings and two, conservative—between the time the justice is nominated and the Senate’s vote.
The resulting scores range from 0 (very conservative) to .5 (moderate) to 1 (very liberal.), and are available at:
http://ws.cc.stonybrook.edu/polsci/jsegal/qualtable.pdf. Segal (and Cover) initially developed them in Jeffrey A.
Segal & Albert D. Cover, Ideological Values and the Votes of Supreme Court Justices, 83 Am. Pol. Sci. Rev. 557
(1989); an updated version appears in Epstein & Segal. supra note 15. We computed the votes cast by justices from
Spaeth, supra note 35
  38
     See supra note 37.
  39
     We derived the votes from Spaeth, supra note 35. Liberal votes are those in favor of defendants in criminal cases;
of women and minorities in civil rights cases; of individuals against the government in First Amendment, privacy,
and due process cases; of unions over individuals and individuals over business in labor cases; and the government
over businesses in economic regulation litigation. Conservative decisions are the reverse.



                                                            6
votes are nearer the top (e.g., Rehnquist) than those who cast a low percentage (e.g., Brennan).




                                                                   80
                                                                                                                                                                                                  Rehnquist
                                                                                                                                                                                                      q
                                                                   70



                                                                                                                                                                                          Alito
                                                                                                                                                                                              q
                                                                                                                                                                             Thomas q
                                                                                                                                                                                          Rehnquist (CJ)
                                                                                                                                                                                               q
                                                                                                                                                                                Roberts qq           q

                                                                                                                                                                                     Burger       Scalia
                Conservative Votes Cast by the Justice (Percent)




                                                                                                                                              O'Connor
                                                                                                                                                 q                                 q
                                                                   60




                                                                                                                                                                                 Powell
                                                                                                                           Whittaker q                   q   Kennedy


                                                                                                Harlan
                                                                                                  q

                                                                                                                                  White
                                                                                                                                      q
                                                                                                         Stewart
                                                                                                            q
                                                                   50




                                                                                                                                                                                          q

                                                                                                                                                                                       Blackmun

                                                                                                                                          q                  Souter
                                                                                                                                     Breyer                    q
                                                                   40




                                                                                                                    q
                                                                                                                                                                         q
                                                                                                                Ginsburg                                              Stevens




                                                                               Fortas
                                                                   30




                                                                                  q


                                                                                  q   Brennan
                                                                                  q                      Goldberg
                                                                                                            q
                                                                              Marshall
                                                                                                            q

                                                                                                         Warren
                                                                   20




                                                                             Very Liberal                                       Moderate                                                          Very Conservative
                                                                                                                            Justice's Ideology



                                                                        Figure 1: The relationship between newspaper editors’ characterizations
                                                                        of justices’ ideology prior to their appointment and the justices’ votes,
                                                                        1953-2005 terms. The superimposed line represents a regression-based
                                                                        prediction of the justices’ votes based on the their ideology. The closer
                                                                        a point to the line, the stronger the association between the justice’s
                                                                        ideology and the justice’s votes. Justices above the line voted more con-
                                                                        servatively than predicted; justices below the line voted more liberally
                                                                        than predicted. The correlation between the justices’ ideology and their
                                                                        votes is .797.40

   If the editors’ ideological assessments do a good job predicting votes, then those justices initially
characterized as conservative, such as Rehnquist, should cast the highest percentage of conservative
votes. Those to the left of center, such as Brennan, should cast the lowest percentage of conservative
votes. Nominees labeled “moderates” by the editors ought be near the center, casting neither many
nor few conservative votes.
    These are the very patterns we observe in Figure 1.41 Indeed, with only scattered exceptions
(e.g., the unexpected liberal voting of Harry Blackmun), press characterizations prior to appoint-
ment turn out to be remarkably good predictors of future voting. To take one example, Ruth
Bader Ginsburg reaches liberal decisions in about 60 percent of the Court’s cases—almost exactly
the percentage we would expect from a justice with her moderately left-of-center political outlook.
Likewise, Antonin Scalia, assessed by all newspaper editors as a conservative at the time of his
nomination, votes precisely as that label would suggest, reaching right-of-center results in almost
seven out of every ten cases he decides. Seen in this way, Ginsburg’s vote against the military
 40
      See supra notes 35, 37, and 39 for information about the data depicted in Figure 1.
 41
      We adapt the discussion in this paragraph from Epstein & Segal, supra note 15.



                                                                                                                                  7
commissions at issue in Hamdan 42 was entirely as predictable as Scalia’s to allow them.


       B   Challenges to Conventional Views

    In light of these findings, it is no wonder why scholars such as Professor Strauss tell us to
disbelieve the possibility of moderation of the part of Chief Justice Roberts. Newspaper editors
characterized him as a conservative at the time of nomination—in the range of a Clarence Thomas
or Warren E. Burger43 —and if the results in Figure 1 are any indication Roberts will vote as
such over the course of his career. Given the new Chief’s presumably well formed views, and his
experience as a constitutional lawyer and appellate judge, to expect otherwise would be foolhardy.
       Or would it?
    Despite the strong consensus over the assumption of stability, several reasons exist to question
it. One prominent challenge comes from a handful of quantitative studies of the justices’ voting.
Rather than summarizing the ideological direction of voting in a single percentage (as we do in
Figure 1), these studies examine the percentage each term. Only by proceeding in this way, the
authors argue, can we detect changes in preferences over time.
    S. Sidney Ulmer’s analysis of the voting patterns of Justices Hugo L. Black and William O.
Douglas is illustrative.44 After plotting their term-by-term support for civil liberties claims, Ulmer
concluded that the two justices evinced substantial change over time: Both began their careers as
relative moderates but grew increasingly willing to support litigants alleging a violation of their
rights—at least until their final years on the bench when their support tapered off a bit (Douglas)
and more than a bit (Black). A replication of Ulmer’s analysis, displayed in Figure 2, seems to
confirm his conclusion that “it cannot be said that Black’s [and Douglas’] support for civil liberty
was stable.”45




  42
     126 S. Ct. 2749 (2006).
  43
     Chief Justice Roberts’ score of .120 is slightly less conservative than Chief Justice Burger’s (.115) and slightly
more liberal than Justice Thomas’s (.160). For more details, see supra note 37 and Figure 1.
  44
     S. Sidney Ulmer, The Longitudinal Behavior of Hugo Lafayette Black: Parabolic Support for Civil Liberties,
1937-1971, 1 Fla. St. U. L. Rev 131 (1973); S. Sidney Ulmer, Parabolic Support of Civil Liberty Claims: The
Case of William O. Douglas, in Courts, Law, and Judicial Processes, (S. Sidney Ulmer, ed., 1981).
  45
     Ulmer, “The Longitudinal Behavior of Hugo Lafayette Black,” supra note 44.


                                                          8
                                                                                                          Justice Black                                                                                                                                                                Justice Douglas
                                                    100




                                                                                                                                                                                                                                              100
                                                                                                                                                                                                                                                                                                                                                q q
  Percentage of Votes Supporting the Rights Claim




                                                                                                                                                                                            Percentage of Votes Supporting the Rights Claim
                                                                                                                                                                                                                                                                                                                                    q
                                                                                                                                  q                                                                                                                                                                                 q                       q                               q
                                                                                                                                                                                                                                                                                                                            q                         q                             q
                                                                                                                                              q       q                                                                                                                                                                         q       q                                       q
                                                                                                                                                                                                                                                                                                                                                          q q           q
                                                    90




                                                                                                                                                                                                                                              90
                                                                                                                                                                                                                                                                                                                                                                                        q
                                                                                                                                          q               q q
                                                                                                                  q                   q                                                                                                                                                                         q                                               q
                                                                                                              q                                                                                                                                                                                         q
                                                                                                                              q                   q                                                                                                                                                                                                                 q
                                                                                                                      q                                                                                                                                                                                                 q
                                                    80




                                                                                                                                                                                                                                              80
                                                                                                                                                                                                                                                                                    q q q
                                                                                                q                         q                                                                                                                                                                                 q
                                                                                    q
                                                                                                                                                                q                                                                                                                                   q
                                                                                                                                                                    q                                                                                                                           q
                                                                                                        q q
                                                    70




                                                                                                                                                                                                                                              70
                                                                q                                                                                                                                                                                   q                   q                   q
                                                                                q
                                                                                                    q
                                                                        q
                                                                                                                                                                                                                                                                    q
                                                                                                                                                                                                                                                            q               q
                                                                                        q
                                                    60




                                                                                                                                                                                                                                              60
                                                                                                                                                                            q

                                                                                                                                                                        q
                                                                            q                                                                                                           q                                                                                       q
                                                                                                                                                                                                                                                                q
                                                          q q                                                                                                                       q
                                                    50




                                                                                                                                                                                                                                              50
                                                                    q                                                                                                           q
                                                                                            q
                                                                                                                                                                                                                                                        q




                                                                1940                                    1950                                  1960                                  1970                                                            1940                               1950                                     1960                                1970

                                                                                                                          Term                                                                                                                                                                              Term



                                                                                            Figure 2: Support for civil liberties claims: The career voting records
                                                                                            of Justices Hugo L. Black and William O. Douglas. This figure reports
                                                                                            the percentage of votes cast each term in which Justices Black (left
                                                                                            panel) and Douglas (right panel) supported defendants in criminal cases;
                                                                                            women and minorities in civil rights cases; and individuals against the
                                                                                            government in First Amendment, privacy, and due process cases. The
                                                                                            superimposed line is a first degree loess smooth with span =0.33.46


    Epstein and her colleagues reached much the same conclusion in their study of the sixteen
justices who sat on the Court for ten or more terms and who began and completed their service
between the 1937 and 1993 terms.47 At least in the area of civil liberties, the authors concluded
that the “preferences of seven justices (Brennan, Burger, Burton, Harlan, Jackson, Marshall, and
Stewart) remained constant over the course of their careers. [But the remaining eleven] changed
in significant linear or nonlinear ways.”48 In other words, most of the justices in their sample grew
increasingly liberal, conservative, or shifted between the two over the course of their career. Espe-
cially noticeable to Epstein and her colleagues, as we show in Figure 3, was Harry A. Blackmun’s
near complete flip, from one of the Court’s most conservative members to among its most consistent
civil libertarians.




  46
     This is an attempt to reproduce Ulmer’s analyses, supra note 44, using data from Lee Epstein, et al., Do Political
Preferences Change? A Longitudinal Study of U.S. Supreme Court Justices, 60 J. Pol. 801 (1998), available at:
http://epstein.law.northwestern.edu/research/prefchange.html.
  47
     Epstein, et al., supra note 46.
  48
     Epstein, et al., supra note 46.


                                                                                                                                                                                            9
                                                                                                                               Justice Blackmun




                                                                                80
                                                                                                                                                                                               q

                                                                                                                                                                                     q
                                                                                                                                                                    q
                                                                                                                                                                                           q




                              Percentage of Votes Supporting the Rights Claim
                                                                                                                                                                                                   q




                                                                                70
                                                                                                                                                                            q

                                                                                                                                                              q
                                                                                                                                                                        q

                                                                                                                                               q




                                                                                60
                                                                                                                                          q
                                                                                                                                                                                q




                                                                                                                       q       q                         q




                                                                                50
                                                                                                        q
                                                                                                                           q
                                                                                                                                    q
                                                                                                                                                     q




                                                                                40
                                                                                            q   q

                                                                                                    q
                                                                                      q
                                                                                                                   q
                                                                                                             q
                                                                                30




                                                                                     1970                   1975                   1980                      1985                   1990               1995

                                                                                                                                              Term




                     Figure 3: Support for civil liberties claims: The career voting record
                     of Justice Harry A. Blackmun. This figure reports the percentage of
                     votes cast each term in which Justice Blackmun supported defendants in
                     criminal cases; women and minorities in civil rights cases; and individuals
                     against the government in First Amendment, privacy, and due process
                     cases. The superimposed line is a first-degree loess smooth with span =
                     0.45.49

    Both Ulmer and the Epstein et al. team speculate on explanations for the trends they observed
but neither puts those explanations to the test. Which may be just as well since their analyses
have their share of problems. Primarily, both studies examine voting records without satisfactorily
attending to the content of the litigation they analyze. As a consequence any observed shifts
in voting could be as much a result of alterations in the cases, as in the justices’ underlying
preferences.50
    To begin to see the problem, consider a justice—call her Justice B —who has served on the
Court for two terms. Suppose that in her first term, Justice B was quite supportive of defendants
in Fourth Amendment cases casting nine out of every ten votes in their favor. In the next term,
however, Justice B voted to support defendants in only one of ten cases. If we looked only at her
votes, we might conclude that our Justice indeed shifted, and shifted to the right: from 90 percent
in favor of defendants to 90 percent against them. But, as Baum points out,51 that conclusion
would be premature. It fails to consider the possibility that the content of the cases varied from
one term to the next—a real possibility, and one with real implications for how we interpret change
(or the lack thereof) on the Court.
    Figure 4, to continue with our example, shows why. Here the horizontal line represents a
single issue dimension, Fourth Amendment search and seizure cases.52 Along that dimension we
have ordered the facts of two cases (as well as three justices) from most liberal (most supportive
  49
      We calculated the percentages depicted in Figure 3 from Spaeth, supra note 35, with value ≤ 6, analu=0,
dec type=1, 6, or 7. See also supra note 39.
   50
      Epstein et al., supra note 46, attempt to account for changes in “issue stimuli” but the approach they use
has its share of problems. Primarily, it is based on a method that assumes preference stability throughout a jus-
tice’s career. For more details, see the critique and reproduction of the Epstein, et al. analysis in Andrew D.
Martin & Kevin M. Quinn, Assessing Preference Change on the U.S. Supreme Court, working paper available at:
http://adm.wustl.edu/papers.php.
   51
      Lawrence Baum, Measuring Policy Change in the United States Supreme Court, 82 Am. Pol. Sci. Rev. 905
(1988).
   52
      We adapt this example Harold J. Spaeth, The Attitudinal Model, in Contemplating Courts (Lee Epstein, ed.


                                                                                                                                        10
of defendants) to most conservative (least supportive of defendants).53 In both cases, 1 (search
warrant) and 2 (no search warrant), the police conducted a search of a home, and in both cases the
searches yielded incriminating evidence. But only in Case 1 did police obtain a warrant. Owing to
the presence of the warrant, Case 1 is more protective of the defendant’s rights than Case 2, and
so we place it to Case 2’s left.
                            Justice A                             Justice B                                Justice C
                             (liberal)                           (moderate)                              (conservative)


                                                  X                                     X
                                                Case 1                                Case 2
                                         (search with warrant)                (search without warrant)


                      Figure 4: Hypothetical Fourth Amendment search and seizure cases and
                      justices in ideological space. In this depiction, justices vote to uphold
                      any search to their left and void any search to the right.54

    Turning to the justices, in Figure 4 we have represented their “most preferred position” or “ideal
point” (i.e., how they would vote in the absence of any internal or external constraints). Here, as
we can see, A is the most liberal, B moderate, and C most conservative. But what conclusion will
Justices A, B, and C reach in the two cases? The answer, under this depiction, is that they will
vote to uphold any search to the left of their ideal point and void any search to the right. In words,
Justice A will vote to strike down the searches in both cases; neither was protective enough of the
defendants’ rights for his taste. Justice C, on the other hand, will vote to uphold both searches;
both, he believes, sufficiently safeguarded the Fourth Amendment. As for our Justice B, she will
agree with C on the warrant case but with A on the warrantless case.
    With this example in mind, we can begin to see the consequences of relying on the percentage
of votes cast, whether in the liberal or conservative direction, to assess preference change among
the justices. Perhaps in Justice B ’s first term, nine of the ten cases involved warrantless searches;
but in her second term, nine of the ten cases involved searches with warrants. If that were the case,
then Justice B ’s preferences did not necessarily move; rather the content of the cases changed—and
changed in a way that made it more difficult for her to cast a liberal vote in her second term relative
to her first.


    C    Importance of Resolving the Debate

   That published studies of ideological movement fail to take into account changes in case content
may render their specific conclusions suspect. Nonetheless, we ignore the potential challenge they
pose to assumption of stability in judicial preferences at our own peril.
    Why? Put simply, and our quibbles with existing studies aside, it is hard to ignore the fact
that by virtually all accounts—from the quantitative to the qualitative, from the historical to
the doctrinal—some justices did move to the left or right during their tenure on the Court, and
moved quite a bit. If the law reviews are any indication, Harry Blackmun appears to be one. His
jurisprudential turn—from a supporter of the death penalty to an opponent, from an advocate
of states’ rights to a proponent of federal power, and from an unwillingness to elevate standards
in sex discrimination litigation to an ardent supporter of women’s rights—are hardly indices of
1995) and Baum, supra note 51, at 905-906.
  53
     To keep the example simple we display only three justices but it easily generalizes to nine, and we could easily
add cases.
  54
     See supra note 52.


                                                                   11
stability.55 And if legal historians are right, Owen Roberts was another. In what commentators in
1937 described as “the switch in time that saved Nine,” Justice Roberts moved from the anti-New
Deal wing of the Court to join President Franklin D. Roosevelt’s four supporters. Whether Roberts’
“switch” was a response to political pressures of the day (i.e., the President’s plan to add one new
seat on the Court for every justice who attained the age of seventy), a growing disenchantment
with the hard-line views of the anti-New Deal justices, or both is still a matter of considerable
debate.56 What is now seemingly settled is that Roberts did move;57 the justice himself implied as
much when asked about his historic shift some years later.58
    But are the Robertses and the Blackmuns anomalies, as the stability assumption would suggest,
or are they the rule, as the Epstein research team might argue? This is the threshold question we
consider, and it is one worthy of sustained attention. Most obviously, it remains an open question.
While the conventional view holds that justices remain committed to the ideological values they
brought to the Court, exceptions are sufficiently numerous and challenges sufficiently compelling
to revisit the received wisdom.
    Second, the debate over preference stability is not only or merely one of academic or theoretical
interest. It is also holds interesting and non-trivial consequences. To see this, we need only imagine
a Court full of Blackmuns, that is, justices who began their career espousing one set of ideological
values and end with another. If that were the true state of the world, rather than the one more
conventionally envisaged, we might reconsider the criteria emphasized during the appointments
process and the possibility of doctrinal change. Both deserve consideration.


          1   The Appointment of Justices

    Beginning with the appointment of justices, we know from historical and contemporary accounts
that most Presidents invest considerable personal energy in selecting the “right” nominee for the
Court.59 In some instances the “right” nominee has little to do with the President’s own ideological
preferences; superior credentials may come into play.60 When the Republican Herbert Hoover chose
Benjamin Cardozo to fill Oliver Wendell Holmes’s seat, the President had no reason to believe that
Cardozo shared his political values. Actually quite the opposite: Cardozo was a Democrat and a
  55
     See, e.g., Ruger, supra note 10; Martha J. Dragich, Justice Blackmun, Franz Kafka, and Capital Punishment,
63 Mo. L. Rev. 853, 853 (1998) (“Over the course of his thirty-five years as a judge, Justice Harry A. Blackmun
seemed to change his views on the death penalty.”; Jeffrey B. King, Now Turn to the Left: The Changing Ideology of
Justice Harry A. Blackmun, 33 Hous. L. Rev. 277, 277 (1996 (“Blackmun very well may have undergone one of
the most marked ideological changes the United States has seen in a public figure during this century.”
  56
     The literature along these lines is vast. As Barry Friedman, The History of the Countermajoritarian Difficulty,
Part Four: Law’s Politics, 148 U. Pa. L. Rev. 971, 1048 (2000), puts it, “since 1937, scholars have debated what
happened and why, combing the historical record in order to ascertain the motives of key players, such as Justice
Owen Roberts, whose possible change of votes in key cases was ‘the switch in time that saved Nine.’”
  57
     As Friedman, supra note 56, 1050, notes “Many scholars simply assume a switch occurred.” These days, only
“legalists argue at least that no switch occurred in response to politics, and perhaps also that no switch at all occurred
in 1937.”
  58
     Justice Roberts’s response was not particularly informative but neither did he deny the move: “Who knows what
causes a judge to decide as he does. Maybe the breakfast he had has something to do with it.” Quoted in Merlo J.
Pusey, Justice Roberts’ 1937 Turnaround, 1983 Y.B. Sup. Ct. Hist. Soc’y 107 (1983).
  59
     See, e.g., David Alistair Yalof, Pursuit of Justices (1999); Henry J. Abraham, Justices, Presidents,
and Senators (1999); Epstein & Segal, supra note 15.
  60
     For more on the goals of Presidents when selecting Justices, see Epstein & Segal, supra note 59; Sheldon
Goldman, Picking Federal Judges (1997).




                                                           12
progressive at that. But “politics aside,” as Silverstein tells us, “there was much to recommend
Cardozo”:

       Cardozo was a highly distinguished jurist, and Hoover was conscious of the potential
       ignominy in being remembered by history as the President who filled the Holmes seat
       on the Court with an unknown. [Besides,] William Borah, the powerful Republican sen-
       ator from Idaho . . . championed Cardozo as the best candidate regardless or residence,
       religion, or party affiliation.61

From Hoover’s perspective, thus, Cardozo was the “best candidate,” not because of his ideology
but because of his stellar credentials.
    In other instances, the best candidate might be the one most able to advance the President’s
or his party’s electoral goals. The moderate Republican Dwight Eisenhower appointed the liberal
Catholic Democrat William J. Brennan, Jr. not because he believed that Brennan shared his
political values or that he was an intellectual heavy weight, but because the President thought he
could gain the support of Catholic voters.62
    Surely other justices have been been appointed for similar reasons but, truth be told, Brennan
and Cardozo are the exceptions. In many, actually most, instances Presidents search long and hard
for nominees who are political allies, not political pawns or prodigious legal minds.63 Why this is
the case is no great mystery: appointing a justice who shares his ideological values and, crucially,
will espouse those values long after he vacates office can result in an unparalleled legacy to the
nation. Some scholars refer to this as “entrenchment,” or the idea that Presidents, in cooperation
with the Senate, can extend their ideological or partisan reach into the federal judiciary not only at
the time of appointment but for the decades to come.64 Perhaps that is why Richard Nixon once
said that, “the most important appointments a President makes are to the Supreme Court of the
United States.”65 He would know. While Nixon left office in 1974, one of his legacies, in the form
of William H. Rehnquist, remained on the Court for three more decades.
    But would Nixon or any other President deem Supreme Court appointments their “most impor-
tant” if ideological drift, even among seemingly rock-solid conservatives (liberals) were the norm,
and not the exception? Unless the President’s goals are more electorally and less ideologically
oriented—not often the case—we suspect not. Presidents typically fret so much about their nom-
inees because they want to ensure a legacy. But if that were a Quixotic project, their time might
be more efficiently spent on advancing other objectives such as appointing justices of superiority
quality, as did Hoover, or those who might improve the party’s electoral fortunes, as did Eisenhower.
    Much the same logic applies to the Senate. Relative to all other positions in the executive
branch, the Senate is far more likely to turn back candidates for the Supreme Court: Since 1789
it has rejected only nine nominees for cabinet posts66 but about one out of every five would-be
  61
     Mark Silverstein, Judicious Choices: The New Politics of Supreme Court Confirmations (1994).
  62
     Epstein & Segal, supra note 59, at 58.
  63
     Epstein & Segal, supra note 59, at 60; Yalof, supra note 59; Brian J. Moraski & Charles R. Shipan, The Politics
of Supreme Court Nominations: A Theory of Institutional Choice and Constraints, 43 Am. J. Pol. Sci. 1069.
  64
     See, e.g., Balkin & Levinson, supra note 15.
  65
     Appears in Transcript of President’s Announcement’s, N.Y. Times, Oct. 22, 1971, at 24.
  66
     Roger B. Taney, Treasury in 1834; Caleb Cushing, Treasury in 1843; David Henshaw, Navy in 1844; James
M. Porter, War in 1844; James S. Green, Treasury in 1844; Henry Stanbery, Treasury in 1868; Charles War-
ren, Justice in 1925; Lewis Strauss, Commerce in 1959; John G. Tower, Defense in 1989. Data available at:
http://www.senate.gov/artandhistory/history/common/briefing/Nominations.htm.


                                                        13
justices.67 What accounts for the comparatively high rate of failure? Surely one factor is the
Constitution’s grant of life tenure for federal judges. With removal for political reasons a near
impossibility,68 senators seem to appreciate the long-term implications of their decisions.
    On the other hand, would life tenure carry as much weight with legislators if their confirmees
voted unpredictably, and unpredictably from one term to the next? We suspect not, and extant
studies are consistent with our suspicion. Most show that a candidates’ ideology is a, if not the,
primary consideration for senators when they cast their votes.69 In fact, the probability of a very
liberal senator voting for a moderately qualified but extremely conservative nominee is under .10;
the likelihood of a very conservative senator voting for that nominee is close to .90.70 To put it
another way, virtually all the senators who cast yea votes for Samuel Alito knew (or at least hoped)
they were voting for a conservative, and hoped they were voting for a conservative for the years to
come. This is the very idea of partisan or ideological entrenchment.71
     A similar calculus, it is worth noting, operates for the many interest groups who lobby against
(for) Supreme Court nominees. From their perspective, spending money to defeat (or support) a
life-long enemy (or ally) on the Court seems a rational course of action—that is, assuming the groups
have accurately predicted the nominee’s ideology and that nominee, as a justice, will continue to
espouse that ideology.72 While the former seems quite possible,73 the latter is precisely what we
question here—and with good reason at that. The many civil rights groups who lobbied against
William Rehnquist may have guessed right—once appointed, he was no friend to their cause—but
they were wrong with regard to David H. Souter.74 In discrimination cases, Justice Souter supports
the plaintiff almost as often as the current Court’s most liberal member, John Paul Stevens.75


          2   The Possibility of Doctrinal Change

    Battles over the appointment of justices are not the only context for which the assumption of
ideological stability has consequences. Another is more doctrinal in nature. It has been common-
place for years, and remains so today, for commentators to promote the idea that legal change can
only come about with membership change or, alternatively, to downplay the possibility of legal
  67
      A list of candidates rejected by the Senate is available at:
http://www.senate.gov/pagelayout/reference/nominations/Nominations.htm.
   68
      For more on this point, see Epstein & Segal, supra note 59, 31-34; Emily Van Tassel, Resignations and Removals:
A History of Federal Judicial Service—and Disservice—1789-1992, 142 U. Pa. L. Rev. 333 (1993); Jack Knight
& Lee Epstein, On the Struggle for Judicial Supremacy 30 L. & Soc’y Rev. 87 (1996).
   69
      See, e.g., Jeffrey A. Segal, Charles M. Cameron, & Albert D. Cover, A Spatial Model of Roll Call Voting:
Senators, Constituents, Presidents, and Interest Groups in Supreme Court Confirmations, 36 Am. J. Pol. Sci. 96
(1992); Charles M. Cameron, Albert D. Cover, & Jeffrey A. Segal, Senate Voting on Supreme Court Nominees: A
Neoinstitutional Model, 84 Am. Pol. Sci. Rev. 525 (1990).
   70
      Lee Epstein, et al. The Changing Dynamics of Senate Voting on Supreme Court Nominees, 68 J. Pol. 296
(2006).
   71
      See, e.g., Balkin & Levinson, supra note 15.
   72
      See, e.g., Gregory A. Caldeira & John R. Wright, Lobbying for Justice: Organized Interests, Supreme Court
Nominations, and the United States Senate, 42 Am. J. Pol. Sci. 499 (1998).
   73
      For evidence of the predictability of nominees during their first term in office, see infra Figure 13.
   74
      Among the civil rights groups testifying against David H. Souter were the National Lawyers Guild, Supreme
Court Watch, and Lambda Legal Defense and Education Fund. See S. Hrg. S.Hrg. 101-1263, Sept. 13, 14, 17, 18,
19, 1990.
   75
      In the 2004 term, Chief Justice Rehnquist voted in favor of civil rights litigants in only 33 percent of the nine
cases in which he participated; those figures for Souter were 83.3 (N=12) and 91.7 for Stevens (N=12). Figures are
from Epstein, et al. supra note 30, Table 6-5.



                                                          14
change in the absence of turnover.76
    Hamdan v. Rumsfeld provides an example but it is hardly the only one. Perhaps the quintessen-
tial case along these lines is Roe v. Wade.77 While the decision has been controversial almost since
the day the Court handed it down, it rises in prominence each time a justice retires. When Lewis
Powell announced his resignation in 1987, journalists emphasized his “crucial” role in retaining the
1973 precedent.78 Two decades later, they said much the same about Sandra Day O’Connor:

        Justice O’Connor’s retirement will not end the court’s majority for Roe, which stands
        at 6 to 3. But her successor could narrow that majority, and open the door to new
        abortion restrictions. For example, the Supreme Court ruled by only 5 to 4 that a
        ”partial birth abortion” ban was unconstitutional; Justice O’Connor’s vote was among
        the five in the majority.79

The assumption here is that Roe cannot go, or even be narrowed in application, unless the Court ex-
periences a turnover in its membership. Ditto for the affirmative action case, Grutter v. Bollinger ;80
death penalty doctrine beginning with Gregg v. Georgia;81 the controversial takings decision in Kelo
v. New London 82 —or, really, any other line of precedent, regardless of its degree of notoriety.
    The accuracy of this view is an open matter, and one we explore more fully in Part IV. The
point here is that its implications are clear. Not only does it work to politicize the confirmation
process—if justices were less predictable over the long term, battles over their appointment ought
diminish as interest groups expend relatively greater resources elsewhere. It also may well affect
the calculus of litigators. If they file petitions only in cases in which their odds of winning are
50-50,83 why bother challenging the right to abortion, the constitutionality of capital punishment,
the taking of private property for economic development, or the use of race in university admissions
in the absence of a membership change? There would be little reason. But should existing justices
experience a change in their jurisprudential outlook, litigation strategy would follow suit, with
petitions continuing to flow in these seemingly closed areas.


       III   Analyzing Preference Change on the Supreme Court, 1937-2005 terms

   Three critical points emerge from our discussion thus far. The assumption of stability (1) is
commonplace (though not unchallenged) and (2) has important implications for the appointment
  76
     See supra notes 25 and 5.
  77
     410 U.S. 113 (1973).
  78
     See, e.g., Linda Greenhouse, Powell: Moderation amid Divisions, N.Y. Times, June 27, 1987, at 32. (“His vote
was crucial in key areas. . . . On abortion he remained committed, with a shrinking majority, to the 1973 precedent
that established it as a constitutional right.”).
  79
     Robin Toner, After a Brief Shock, Advocates Quickly Mobilize, N.Y. Times, July 2, 2005, at 1A. The same article
reports the reaction of interest groups: “At the abortion rights group Naral Pro-Choice America, organizers were
sending e-mail alerts to 800,000 activists within 15 minutes after the announcement of Justice Sandra Day O’Connor’s
resignation. ‘Don’t let Bush take away your choice!’ they declared.”
  80
     539 U.S. 306 (2003).
  81
     428 U.S. 153 (1976).
  82
     126 S. Ct. 24 (2006).
  83
     The Priest-Klein model of litigation predicts that plaintiffs only will go into litigation if they believe that they
have roughly a fifty percent chance of winning. George L. Priest & Benjamin Klein, The Selection of Disputes for
Litigation, 13 J. Legal Stud. 1, 45 (1984). This prediction is contingent on the decision standard, the parties
uncertainty of estimating case quality, and the degree of stake asymmetry across the parties.


                                                          15
and work of the justices but (3) is tricky to assess empirically. The primary difficulty is how to
solve the vexing problem of variation in case content, and how to solve it on a large-scale basis.84
    These questions have perplexed scholars for decades but, fortunately for us, Martin and Quinn,
two coauthors of this article, have devised a satisfactory solution.85 Using data derived from the
votes cast by the justices and a Bayesian modeling strategy, they have generated term-by-term
ideal point estimates for all the justices appointed since the 1937 term—estimates that attend to
variation in case content. In other words, using the Martin-Quinn approach we can offer intra-
justice comparisons (e.g., is Justice Souter more liberal now than he was in 1992?) without having
to consider whether the changes we observe are the result of differences in the content of cases or
changes in the justice’s revealed preferences.86
    Not surprisingly, the products of the Martin-Quinn method–i.e., their ideal point estimates—
have received a good deal of play both in the popular press and in scholarly journals.87 We too
have deployed them in a study of the median justice on the Supreme Court;88 and Ruger, along
with Martin & Quinn,89 have even invoked them to analyze change on the Court, though for a
limited set of justices. Hence, in an effort to conserve space, we direct readers interested in learning
more about the Martin-Quinn procedures to these other sources, as well as to web sites housing the
data.90 The important point to underscore here is that their estimates overcome major obstacles
of the past and thus allow us to make high-quality inferences about justices’ voting over time.
    What are those inferences? What can we learn about preference change on the Court from the
Martin-Quinn estimates? Are the justices as stable as most commentators seem to assume? Or
is change the rule, not the exception? Figures 5, 7, 9, and 12 address these questions,91 and the
  84
     Some scholars, most notably, Jeffrey A. Segal, Measuring Change on the Supreme Court: Examining Alternative
Models, 29 Am. J. Pol. Sci. 461 (1985), have developed area-specific solutions; in Segal’s case, Fourth Amendment
search and seizure litigation. But we know of no work that satisfactorily tackles the problem across the range of legal
areas.
  85
     Andrew D. Martin & Kevin M. Quinn, Dynamic Ideal Point Estimation via Markov Chain Monte Carlo for the
U.S. Supreme Court, 1953-1999, 10 Political Analysis 134 (2002).
  86
     Because, as we mention in the text, the Martin-Quinn method has been described elsewhere, see Martin &
Quinn, supra note 85 and infra note 87, suffice it to note here that their method simultaneously provides comparable
estimates of ideal points and cut points (the midpoint between the status quo policy and the potential policy under
review; for more details, see infra note 157) by (1) exploiting the overlapping service records of justices and (2)
assuming that model parameters governing the cut points are drawn from a common distribution. Overlapping
service records allow for model-based comparisons of justices who never served together. For instance, Martin and
Quinn use the fact that Chief Justice Warren served with Justice Brennan who served with Justice Scalia to place
the ideal points of Chief Justice Warren and Justice Scalia (who never served together) on a comparable scale. The
approach achieves intertemporal comparability by assuming that ideal points can onlychange smoothly through time
and that any voting records that are consistent with all justices moving equally to the right or equally to the left are
the result of changes in the cut points rather than Court-wide changes in preferences.
  87
     See, e.g., Ruger supra note 10; Barry Friedman & Anna L. Harvey, Electing the Supreme Court, 78 Ind. L.J. 123
(2003).; Paul J. Wahlbeck, The Chief Justice and the Institutional Judiciary: Strategy and Constraints on Supreme
Court Opinion Assignment, 154 U. Pa. L. Rev. 1729 (2006). For media reports of their research see, e.g., the Wall.
St. J., February 1, 2006, at A4; Wash. Post, November 1, 2004, at A19.
  88
     Andrew D. Martin et al., The Median Justice on the United States Supreme Court, 83 N.C. L. Rev. 1275
(2005).
  89
     Ruger, supra note 10; Andrew D. Martin & Kevin M. Quinn, Assessing Preference Change on the U.S. Supreme
Court, unpublished ms. available at http://adm.wustl.edu/papers.php
  90
     http://epstein.law. northwestern.edu/research/ideodrift.html contains all the data used in this article.
http://adm.wustl.edu/supct.php also houses the Martin-Quinn estimates, as well as annual updates. See also supra
note 85 and supra note 86.
  91
     More specifically, Figures 5, 7, 9, and 12 depict the estimated ideal points over time for each justice. To reach
conclusions about ideological drift, we examined the preference change profiles for each justice. Examples of these



                                                          16
answer could not be clearer. Of the twenty-six justices who served on the Court for ten or more
terms since 1937, all but four exhibit ideological drift over the course of their tenure.92 Twelve
moved to the left;93 seven to the right;94 and three in more exotic ways.95


    A    Trending to the Left

    We begin, in Figure 5, with the twelve justices who grew more liberal during their tenure.
That we find Harry A. Blackmun among this group is hardly surprising. Blackmun may have
once quipped that it was not he but the Court that changed, but he also said, “I suspect that
when one goes on the Supreme Court of the United States his constitutional philosophy is not fully
developed. . . And if one didn’t grow and develop down there I would be disappointed in that person
as a Justice.”96 Quite clearly, and just as most analysts would have predicted, the latter is more
descriptive of his career.




profiles appear in Figures 6, 8, and 10. The full complement is available on the project’s web site, as is a description
of the procedures we used to identify significant drift.
  92
     The four are Justices Breyer, Murphy, Stewart, and Thomas. See infra Part IIIC.
  93
     Justices Blackmun, Brennan, Clark, Ginsburg, Kennedy, Marshall, O’Connor, Powell, Souter, and Stevens, and
Chief Justices Rehnquist and Warren. See infra Part IIIA.
  94
     Justices Black, Burton, Frankfurter, Jackson, Reed, Scalia, and White. See infra Part IIIB.
  95
     Chief Justice Burger, and Justices Douglas and Harlan. See infra Part IIIC.
  96
     Ruger, supra note 10, at 1212 quoting Justice Harry A. Blackmun, Speech at Louisiana State University Law
Center, Summer Program (July 6-9, 1992, Aix-en-Provence, France).


                                                          17
                                     Justice Blackmun                            Justice Brennan                                         Justice Clark




                         6




                                                                   6




                                                                                                                       6
                         2




                                                                   2




                                                                                                                       2
                         -2




                                                                   -2




                                                                                                                       -2
                         -6




                                                                   -6




                                                                                                                       -6
                              1970     1980   Term



                                                            1990        1955   1965     Term



                                                                                               1975      1985               1945            1955   Term



                                                                                                                                                                    1965




                                     Justice Ginsburg                           Justice Kennedy                                        Justice Marshall
                         6




                                                                   6




                                                                                                                       6
                         2




                                                                   2




                                                                                                                       2
                         -2




                                                                   -2




                                                                                                                       -2
                         -6




                                                                   -6




                                                                                                                       -6
                              1990            Term




                                                     2000               1985           1995
                                                                                        Term




                                                                                                                2005        1965       1975        Term




                                                                                                                                                             1985            1995




                                     Justice O'Connor                            Justice Powell                                    (Chief) Justice Rehnquist
                         6




                                                                   6




                                                                                                                       6
                         2




                                                                   2




                                                                                                                       2
                         -2




                                                                   -2




                                                                                                                       -2
                         -6




                                                                   -6




                                                                                                                       -6




                              1980     1990   Term



                                                            2000        1970           1980
                                                                                        Term



                                                                                                                1990        1970     1980          Term



                                                                                                                                                          1990        2000




                                      Justice Souter                             Justice Stevens                                     Chief Justice Warren
                         6




                                                                   6




                                                                                                                       6
                         2




                                                                   2




                                                                                                                       2
                         -2




                                                                   -2




                                                                                                                       -2
                         -6




                                                                   -6




                                                                                                                       -6




                              1990            Term




                                                     2000               1975    1985    Term




                                                                                                  1995          2005        1950               1960Term




                                                                                                                                                                             1970




                      Figure 5: Estimated ideal points of the justices who served ten or more
                      terms between 1937 and 2005, and trended left. The vertical axis in
                      all plots is the justice’s estimated ideal point. Higher values are more
                      conservative. The dots are the estimated ideal points and the vertical
                      error bars in each plot are 95% credible intervals. 97

    Also less than startling in light of contemporary commentary is Justice O’Connor’s behavior.98
While the data displayed in Figure 5 seem to show relatively consistent preferences over time,
further analysis depicts a justice who in fact trended to the left. To see this consider Figure 6, in
which we display the probability that O’Connor was more conservative in any given term than in
all others. Within the figure, the baseline term is on the vertical axis and the comparison term
  97
     For more details on the estimated ideal points, see supra notes 85 and 86.
  98
     See, e.g., Joan Biskupic, O’Connor Not Confined by Conservatism, USA Today, June 24, 2004, at 4A (“Although
O’Connor usually votes with the courts conservative wing, she increasingly has sided with liberals in significant
cases that have been decided by 5-4 votes. It’s led some conservative observers to wonder whether O’Connor, at 74,
is turning to the left.”); Charles Rothfeld, The Court on Balance; By Sometimes Leaning Left, Justice O’Connor
Centers the Supreme Court, Legal Times, July 12, 2004, at 52 (“The liberals dominated in the eight civil cases decided
by 5-4 votes, winning six of them. O’Connor voted with the liberal majority in four of these cases.”).



                                                                                      18
is on the horizontal axis. E.g., if O’Connor was equally as conservative in 2000 as she was say,
in 1985 then beginning at the 1985 mark on the vertical axis and moving to the right, we would
expect to see neither bright red (indicating that she grew significantly more liberal) nor bright blue
(indicating a significant move to the right) but a blackish color. That we do not see. Beginning in
the early 1990s, only red appears, indicating a significant turn to the left relative to her voting in
the 1980s. Of course scholars and journalists not only took note of this trend but also speculated
on its doctrinal consequences (a subject to which we return in Part IV).
    While the increasing liberalism exhibited by Justice O’Connor, not to mention Justices Black-
mun and Stevens, may come as a surprise to very few, we cannot say precisely the same of the
others depicted in Figure 5, especially David H. Souter, Anthony Kennedy, and two of the three
most recent chief justices, Rehnquist and Warren. When George H.W. Bush selected Souter to
serve on the Court in 1990, the President had any number of reasons to believe he was appointing
a justice who would cast consistently conservative votes, whether over abortion, prayer in school,
criminal rights, or affirmative action. This is not to say that Bush could have opted for an even
more reliable solid conservative; in fact he considered several, including Edith Jones.99 But by
most accounts Souter was reliable enough.100 Even newspaper editors (and editors of all ideo-
logical stripes at that) thought as much. Before Souter joined the Court, they deemed him even
more conservative than two of Ronald Reagan’s appointees, Sandra Day O’Connor and Anthony
Kennedy at the time of their nominations.101
    Figure 5 reveals that the President and the editors were not wrong—at least not initially. For
the 1990 term Souter’s ideal point estimate places him far closer to, say, Justices O’Connor and
Kennedy than those on the extreme left (Justices Blackmun, Stevens, Marshall). Actually during
his first two terms, Souter was the Court’s likely median, or swing, justice.102 That Souter’s ideal
point is now closer to the the most liberal member of the Court (Stevens) than to the middle
(Kennedy) has not been been missed by Court observers.103 But the extreme leftward movement is
notable. Indeed, Souter is the new Blackmun; that is, a justice who, as Figure 6 shows, has grown
strikingly more liberal with nearly each passing term.
    Pointing to his pivotal role in establishing liberal majorities in Lawrence v. Texas 104 and Roper
v. Simmons,105 some analysts have asserted much the same about Anthony Kennedy, that he has
moved significantly to the left.106 While it is true, as Figure 6 shows, that Justice Kennedy drifted
  99
     For an account of the Souter nomination, see Yalof, supra note 59.
 100
     On her blog, at http://www.anncoulter.com/cgi-local/article.cgi?article=67, the conservative commentator, Ann
Coulter, provides quote after quote attesting to Souter’s conservative credentials at the time of his appointment. E.g.,
Newt Gingrich claimed that “Virtually every conservative who knows him trusts him and thinks he’s a competent
guy.” The National Right to Life’s John Willke said, “(He) seems to be a judicial conservative, what we call a
constitutional constructionist.. . . That’s satisfactory with us, if that’s true.”
 101
     See supra Figure 1 and supra note 37.
 102
     Using the Martin & Quinn scores, Martin, et al., supra note 88, have calculated the justice most likely to have
been the median for each term. Souter is that justice for the 1990 and 1991 terms, though the probability that he
was the median is reasonably weak (.48 in 1990 and .34 in 1991). Data appear in Martin, et al., supra note 88.
 103
     E.g., Richard H. Fallon, Jr., The “Conservative” Path of the Rehnquist Court’s Federalism Decision, 69 U. Chi.
L. Rev. 429, 494 (deeming Souter a “liberal”); Erwin Chemerinsky, The Rehnquist Court and the Death Penalty,
94 Geo. L.J. 1367, 1368 (2006) (placing Souter in the Court’s “liberal wing”); Scott P. Johnson & Robert M.
Alexander, The Rehnquist Court and the Devolution of the Right to Privacy, 105 W. Va. L. Rev. 621, 645 (2003)
(Justice Souter has “surprised many conservatives with his moderate to liberal voting record.”).
 104
     539 U.S. 558 (2003).
 105
     543 U.S. 551 (2005)
 106
     e.g., Thomas Sowell, Justice Kennedy’s New Move Left: Soft on Crime, Hum. Events, Aug. 18, 2003; Lawrence
Friedman, The Limitations of Labeling: Justice Anthony M. Kennedy and the First Amendment, 20 Ohio N.U. L.


                                                          19
                                 Justice Kennedy                                                                    Justice O'Connor




                                                                                               2005
         2005




                                                                                               2000
         2000




                                                                            Reference Term

                                                                                               1995
         1995




                                                                                               1990
         1990




                                                                                               1985
                         1990           1995            2000            2005                                 1985     1990    1995          2000   2005
                                Justice Rehnquist
                                  Comparison Term
                                                                                                                     Justice Souter
                                                                                                                      Comparison Term




                                                                                               2005
         2000
         1995




                                                                                               2000
                                                                            Reference Term
         1990
         1985




                                                                                               1995
         1980
         1975




                                                                                               1990




                       1975    1980    1985     1990    1995     2000                                 1990          1995             2000          2005
                                     Justice Warren
                                      Comparison Term                                                                 Comparison Term
         1965
         1960
         1955




                       1955              1960                 1965

                                                                                 Legend




                0.00          0.10       0.20          0.30          0.40                    0.50        0.60        0.70     0.80          0.90    1.00

                                                                 Probability(More Conservative)



Figure 6: Estimated preference change profiles for five left-trending jus-
tices. The baseline term is on the vertical axis, and the comparison term
is on the horizontal. E.g., suppose we are interested in whether Justice
O’Connor is more conservative in terms subsequent to 1985 than in the
1985 term. Begin at the 1985 mark on the vertical axis, and read the
colors across from left to right. Then consult the legend to see how
the probabilities are encoded in the colors, from bright red (indicating
that the justice is significantly more liberal) through bright blue (indi-
cating that the justice is significantly more conservative). As for Justice
O’Connor, the bright red tells us that in many terms after 1985 she was
significantly more liberal than she was in that term.




                                                                                       20
to the left early on his career—he is now significantly more liberal now than he was in, say, 1988—
since the early 1990s his ideal point has remained flat. Given Kennedy’s crucial role as the pivotal
justice on the current Court, this is a finding replete with interesting implications, and we consider
them in some detail in Part V.
    Equally interesting are the patterns of the two chief justices depicted in Figures 5 and 6, Warren
and Rehnquist. Juxtaposed against each other we observe change, though the trends differ. At
first blush, Warren’s revealed preferences appear quite stable (see Figure 5). The more detailed
analysis depicted in Figure 6 confirms a high degree of consistency, though with two important
exceptions: his first two terms on the Court. Note the bright red color at the bottom of Warren’s
panel, revealing that the Chief Justice became far more likely to exhibit liberal preferences as time
marched on. In this way, he resembles Souter, another justice whose behavior altered after his
early years on the Court. Unlike Souter, however, Warren did not continue to waiver: By 1955 he
became a consistent liberal, neither veering much to the left or right thereafter.
    By contrast come Rehnquist’s ideal point estimates—estimates that are perhaps the most unex-
pected of all the liberal-trending justices. When Nixon appointed Rehnquist to the Court in 1971,
newspaper editors and scholars alike agreed on his ideological propensities: Without doubt, they
said, he was a solid, if not an extreme, conservative. When Ronald Reagan elevated Rehnquist
to Chief Justice in 1986, the refrain was similar: the New York Times declared him a member of
the Court’s “extreme right wing.”107 Even when Rehnquist died in September of 2005, the press
continued to label him the “architect of [a] conservative court.”108 In short, for over thirty years
Rehnquist was tagged as one the Court’s most reliable right-of-center votes.
    The story emerging from Figures 5 and Figure 6 is more complicated. To be sure, when he
joined the Court, Rehnquist’s ideal points placed him as the most extreme (conservative) justice.
In fact during the mid-1970s he was to the right of where Clarence Thomas—today’s most extreme
conservative—is now.109 But, when Rehnquist was promoted to Chief Justice and Scalia joined
the Court, Rehnquist begin to drift left. Note the bright red coloring in Figure 6, indicating that
in every term between 1986 and his death in 2004, Rehnquist’s preferences were significantly more
liberal than in 1985.
   Of course, this is not to say that Rehnquist swung as far to the left as Harry Blackmun; he did
not. On the other hand, in the Chief Justice’s last term in office his ideal point estimate is closer to
the centrist Kennedy’s than to the extreme conservative position that he once held or that Scalia
and Thomas now anchor.110


    B   Trending to the Right

    That Rehnquist trended to the left is interesting if only because our results refute portrayals of
his voting as monolithically conservative over the course of his tenure. Of even greater interest, of
course, is whether his movement affected his decisions. Was it the case that Rehnquist was so far
to the right that his liberal turn simply made him a less extreme conservative or are traces of the
Rev. 225 (1993) (Kennedy has become more liberal in some legal areas).
 107
     Toward a Rehnquist Court, June 18, 1986, at 34A.
 108
     E.g., Linda Greenhouse, William H. Rehnquist, Architect of Conservative Court, Dies at 80, September 5, 2005,
at 16A.
 109
     E.g., Rehnquist’s estimated ideal point in 1975 is 4.22 versus Thomas’s, thirty years later in 2004, of 3.45.
 110
     See infra Figure 19.



                                                       21
turn reflected in his jurisprudence?
    We address that important question momentarily. For now, consider those justices who, in
contrast to Rehnquist and the others, trended to the right. Falling into this category, as we can see
in Figure 7, are Justices Hugo Black, Harold Burton, Felix Frankfurter, Robert Jackson, Stanley
Reed, Antonin Scalia, and Byron White.




                                                 22
                                            Justice Black                                                  Justice Burton




                        6




                                                                                               6
                        2




                                                                                               2
                        -2




                                                                                               -2
                        -6




                                                                                               -6
                             1935   1945           1955
                                                    Term




                                                                     1965          1975             1945          Term




                                                                                                                         1955




                                       Justice Frankfurter                                                 Justice Jackson
                        6




                                                                                               6
                        2




                                                                                               2
                        -2




                                                                                               -2
                        -6




                                                                                               -6
                             1935      1945         Term




                                                                  1955             1965             1940          Term




                                                                                                                         1950




                                            Justice Reed                                                    Justice Scalia
                        6




                                                                                               6
                        2




                                                                                               2
                        -2




                                                                                               -2
                        -6




                                                                                               -6




                             1935           1945    Term




                                                                         1955                       1985        1995
                                                                                                                  Term




                                                                                                                                2005




                                            Justice White
                        6
                        2
                        -2
                        -6




                             1960    1970           Term




                                                           1980             1990




                     Figure 7: Estimated ideal points of the justices who served ten or more
                     terms between 1937 and 2005, and trended right. The vertical axis in
                     all plots is the justice’s estimated ideal point. Higher values are more
                     conservative. The dots are the estimated ideal points and the vertical
                     error bars in each plot are 95% credible intervals.111

    That even seven justices drifted to the right may, in and of itself, be notable. When commenta-
tors describe ideological change on the Court, they often speak of those who grew more liberal, the
Blackmuns and the O’Connors. Some analysts have concluded that justices, if they move, nearly
always turn left.112 To be sure, this is the dominant pattern among the justices we study—eleven
of the twenty-six became significantly more liberal at some point in their career (including all but
 111
     For more detail on the estimated ideal points, see supra notes 85 and 86.
 112
     E.g., Jon D. Hanson & Adam Benforado, The Drifters: Why the Supreme Court Makes Justices More Liberal,
Boston Rev. (January/February 2006) (“While there have been a number of relatively reliable conservative justices
over the years . . . the tendency in recent decades to drift leftward has been strong enough to gain both popular
and scholarly attention.”); Lawrence Baum, Judges and Their Audiences (2006, 149) (“Among the nine
Republicans who moved to Washington to join the Supreme Court, there were clear and substantial increases in
liberalism in four and more limited or ambiguous increases for three others.”).



                                                                                          23
three members of the current Court)—but movement to the right is no small phenomenon in our
data set.
    Even so, close Court watchers will likely see no big surprises in Figure 7. Four decades ago,113
the political scientist Harold J. Spaeth debunked the oft-repeated claim that Frankfurter was one
of the “most ardent and consistent advocates of judicial restraint.”114 After demonstrating that
Frankfurter’s judicial restraint was “thoroughly subordinated” to his conservative values Spaeth’s
response was succinct: “Ardent? Perhaps. Consistent. No.”115 Our results here show that Frank-
furter was no more consistent with regard to his ideology. He began his career, in the 1938 term, as
a slightly left-of-center justice, closer to the term’s likely median Chief Justice Stone than to either
of the extremes, Hugo Black on the left and James McReynolds on the right. Virtually from the
start of his second term, however, Frankfurter appears to drift right—a trend Figure 8 confirms.
Note the bright blue at the bottom of his panel, indicating a near 1.0 probability that he was more
conservative in later terms relative to his first few years on the Court. By the conclusion of his
tenure, Frankfurter was second only to John Harlan as the Court’s most extreme conservative voter;
and he actually ended his service more firmly planted on the right than Chief Justice Rehnquist.




  113
      Harold J. Spaeth, The Judicial Restraint of Mr. Justice Frankfurter—Myth or Reality, 8 Am. J. Pol. Sci. 22
(1964).
  114
      Henry J. Abraham, Line-Drawing between Judicial Activism and Restraint: A Centrist Approach and Analysis,
in Supreme Court Activism and Restraint (Stephen C. Halpern & Charles M. Lamb, eds. 1982) (our emphasis).
  115
      Spaeth’s claim grows out of Spaeth, supra note 113, but appears in Segal & Spaeth, supra note 28, at 409, n. 6.


                                                         24
                                                   Justice Frankfurter                                                 Justice Black




                                                                                                         1970
                              1960
                              1955




                                                                                                         1960
                                                                                      Reference Term
                              1950




                                                                                                         1950
                              1945
                              1940




                                                                                                         1940
                                            1940     1945      1950   1955     1960                             1940     1950          1960      1970
                                                          Justice White
                                                      Comparison Term                                                  Comparison Term

                              1985
                              1975
                              1965




                                             1965            1975       1985
                                                                                                Legend




                                     0.00          0.10      0.20     0.30     0.40                    0.50     0.60   0.70     0.80      0.90    1.00

                                                                             Probability(More Conservative)



                     Figure 8: Estimated preference change profiles for three right-trending
                     justices. The baseline term is on the vertical axis, and the comparison
                     term is on the horizontal. For more details on how to interpret the figure,
                     see Figure 6.



    Likewise, despite “categorical denial[s] that he had changed his constitutional philosophy,”116
Hugo L. Black’s movement to the right was not missed by some commentators. As James F.
Simon once wrote, Black’s “increasingly brittle, unmistakably conservative tilt,” actually proved
embarrassing to many of his admirers.117 The results depicted in Figures 7 and 8 confirm the
rightward trend throughout Black’s career and especially since the 1960s. Note that in every term
after 1960 the probability that Black was more conservative bordered on 1.0.
    Justice Black served on the Court between the 1938 and 1970 terms. The other justices displayed
in Figure 7 are of an equally old vintage, all completing their tenures prior to Black—with two
notable exceptions: Antonin Scalia and Byron White. Justice Scalia is the only member of the
current Court to have grown consistently more conservative with time, an interesting pattern we
investigate more closely in Part V. White too moved to the right but his revealed preferences
are more intriguing (see Figure 8). Compared with the early 1960s, he grew significantly more
conservative at the start of the Burger Court era; then relative to the early 1980s, he once again
took turn to the right during the onset of the Rehnquist Court.
   Some analysts claim that that these changes are illusory. They say that White “ardently
supported individual rights over the claimed rights of the states to abridge citizens’ liberties” but
 116
    Gerald T. Dunne, Hugo Black and the Judicial Revolution (1977), 413.
 117
    James F. Simon, Judging the Justices, 49 Stan. L. Rev. 173, 176 (1996). See also Jed Handelsman Shugerman,
A Six-Three Rule, 37 Ga. L. Rev. 893, 922 (2003), (“From 1962 to 1969, the liberals gained their fifth vote, first
with Arthur Goldberg and then Abe Fortas. As Justice Black became more conservative, Thurgood Marshall replaced
Clark, and the Court retained its reliable five-vote liberal majority.”).




                                                                                                       25
that “on issues of law enforcement . . . [he] voted conservatively.”118 Others, however, contend that
Justice White was, in fits and starts, more conservative over time.119 Two authors of this article
even speculated that his level of conservatism varied by the President in office.120 Whether this
hypothesis holds we cannot say without more analysis. But our data do lend support to claims
about Justice White’s rightward drift at various points throughout his career.


    C    The Remaining Justices

    Exhibiting even more exotic patterns than Justice White are William O. Douglas, John Harlan,
and Warren E. Burger. As we can see in Figures 9 and 10, both Harlan and Douglas made early
and significant moves to the right, followed by change to the left. The depth of their ideological
commitment was quite distinct: For his last two decades on the Court, Douglas was its most
liberal member; Harlan’s liberalism never surpassed, say, O’Connor’s. Nonetheless, their trends are
similar.
    Relative to Douglas, Chief Justice Burger’s ideal point appears to be consistently conservative,
as virtually all previous analyses suggest.122 As we can see in Figure 10, however, the Chief’s
ideological inclinations were more volatile than many contend. Relative to the early 1970s, Burger
was significantly less likely to reveal conservative preferences in the late 1970s; but compared with
the early 1980s he was significantly more right-of-center at the end of his tenure, that is, by the
late Reagan years.
    Seen in this way our analysis presents something of a challenge to analyses that cluster Burger
and his successor, Rehnquist—or, at least the Courts they led.123 While it is true, as we show
in the left panel of Figure 11, that both Chiefs were more conservative than the Court’s median
(typically Justice White), Burger was considerably more moderate than Rehnquist. In fact, though
much has been made, and rightfully so, about the growing rift between the boyhood friends from
  118
      D. Wes Sullenger, Burning the Flag: A Conservative Defence of Radical Speech and Why It Matters Now, 43
Brandeis L.J. 597, 619 (2005).
  119
      David O. Stewart, White to the Right?, A.B.A. J. July 1990, at 40 (arguing that Justice White assumed more
conservative positions over time); David D. Meyer, Justice White and the Right of Privacy: A Model of Realism and
Restraint, 52 Cath. U.L. Rev. 915 915 (2003) (“popular accounts” portray “White as the increasingly conservative
curmudgeon on matters of individual liberty, wielding traditional morality or his own crusty predilections to repel the
claims of modern society”); Jeffrey Rosen, The Next Justice, New Republic, Apr. 12, 1993, at 21, 24 (White moved
to the ideological right).
  120
      Lee Epstein, Jack Knight, & Andrew Martin, The Supreme Court as a Strategic National Policy Maker, 50
Emory L. J. . 583 (2001).
  121
      For more detail on the estimated ideal points, see supra notes 85 and 86.
  122
      Craig M. Bradley & Joseph L. Hoffmann“Be Careful What You Ask For”: The 2000 Presidential Election,
the U.S. Supreme Court, and the Law of Criminal Procedure, 76 Ind. L.J. 889, 893 (2001) (deeming Burger a
“consistent conservative”); Eric M. Freedman, A Lot More Comes into Focus When You Remove the Lens Cap, 81
Iowa L. Rev. 883, 968 (1996) (“Chief Justice Burger and his conservative colleagues clearly wished to tighten
legal controls on obscenity”); Mark V. Tushnet, The Supreme Court and Race Discrimination, 1967-1991: The View
from the Marshall Papers, 36 Wm & Mary L. Rev. 473 (“Chief Justice Burger, of course, had more conservative
instincts than many of the Justices on the Court when he arrived.”).
  123
      Charles M. Lamb, Chief Justice Warren E. Burger: A Conservative Chief for Conservative Times, in The
Burger Court: Political and Judicial Profiles 129, 132 (Charles M. Lamb & Stephen C. Halpern eds., 1991);
Rebecca E. Zietlow, To Secure These Rights: Congress, Courts and the 1964 Civil Rights Act 57 Rutgers L. Rev.
945, 990 (2005) (“the Court took a more conservative turn under the leadership of Burger and Rehnquist”). But see
Frank B. Cross & Stefanie Lindquist, The Decisional Significance of the Chief Justice, 154 U. Pa. L. Rev. 1665,
1685 (2006) (“Justice Rehnquist is regarded as more conservative than Justice Burger).




                                                          26
                                  Justice Burger                                         Justice Douglas




            6




                                                                      6
            2




                                                                      2
            -2




                                                                      -2
            -6




                                                                      -6
                 1965                 1975
                                        Term




                                                          1985             1935   1945        1955
                                                                                                Term




                                                                                                           1965   1975




                                  Justice Harlan
            6
            2
            -2
            -6




                   1955                 1965
                                        Term




                                                          1975




                          Figure 9: Estimated ideal points of justices serving ten or more terms
                          between 1937 and 2005 and who moved to the left and right. The vertical
                          axis in all plots is the justice’s estimated ideal point. Higher values are
                          more conservative. The dots are the estimated ideal points and the
                          vertical error bars in each plot are 95% credible intervals. 121


Minneapolis, Blackmun and Burger,124 the two generally remained ideologically closer than Burger
and Rehnquist. Only in Burger’s last two terms in office, as the right panel of Figure 11 makes
clear, were the present and future chiefs as aligned in ideological space as the “Minnesota twins,”
Blackmun and Burger.
    This finding may signal the need to revisit legal commentary of the day (and even today)
equating the ideological tendencies of Chief Justices Burger and Rehnquist. Even more in need of
revisiting is the account that motivated our project here: that justices generally do not change in
their ideological outlook over time. According this account, it is only a handful of anomalies, the
very few Blackmun’s, who exhibit fluctuation; the balance remain stable. As it turns out and as
we show in Figure 12, precisely the opposite is true. Only four of the twenty-six justices serving
since 1937 remained relatively stable.
    Two of the four, Frank Murphy and Clarence Thomas, were consistently extreme—Murphy on
the left and Thomas on the right. Justice Murphy may never have been the Court’s most liberal
member during any term on which he served (Black or Douglas held that distinction), but in
several he came quite close.127 On today’s Court, Murphy would be approximately slightly to right
of Souter but to the left of Breyer. As for Thomas, ever since he joined the Court in the 1991 term
he and Scalia have vied for the most conservative spot. But these days, even with Scalia’s turn to
 124
     See, especially, Greenhouse, supra note 12.
 125
     For more detail on the estimated ideal points, see supra notes 85 and 86.
 126
     For more detail on the estimated ideal points, see supra notes 85 and 86.
 127
     E.g., in the 1947 term, Black was the most liberal, with an estimated ideal point at -1.727; Murphy’s was -1.640.


                                                                 27
                                                                                                     Justice Douglas                                                                                 Justice Harlan




                                                                                                                                                              1970
                                                                                  1970




                                                                                                                                                              1965
                                                                                                                                           Reference Term
                                                                                  1960




                                                                                                                                                              1960
                                                                                  1950




                                                                                                                                                              1955
                                                                                  1940
                                                                                           1940           1950          1960      1970                                                      1955           1960      1965            1970
                                                                                  1985
                                                                                  1980
                                                                                  1975
                                                                                  1970                Justice Burger
                                                                                                      Comparison Term                                                                                 Comparison Term




                                                                                            1970           1975           1980       1985
                                                                                                                                                     Legend




                                                                                         0.00      0.10          0.20      0.30     0.40                    0.50                            0.60          0.70    0.80      0.90          1.00

                                                                                                                                  Probability(More Conservative)



                                                                   Figure 10: Estimated preference change profiles for three right- and left-
                   d 1
                  i5 a E
                r m 89R
tnioP laedIsdsitmmhel1-
            nu aag 7sB
            t J neniteT
                   c 0
       ecit ue uqk0ruM
tnioP laedI detamitsE  2
                       4
                       3
                                                                   trending justices. The baseline term is on the vertical axis, and the
                                                                   comparison term is on the horizontal. For more details on how to read
                                                                   the figure, see Figure 6.
                                                   4




                                                                                                                                                                                       4
                                                   3




                                                                                                                                                                                       3
                           Estimated Ideal Point

                                                   2




                                                                                                                                                                                       2
                                                                                                                                                               Estimated Ideal Point
                                                   1




                                                                                                                                                                                       1
                                                   0




                                                                                                                                                                                       0
                                                   -1




                                                                                                                                                                                       -1




                                                        1970               1975                            1980                          1985                                                      1970                            1975                     1980                    1985

                                                                                    Term                                                                                                                                                         Term

                                                               Rehnquist                  Burger                        Median Justice                                                                            Burger                         Blackmun          Median Justice




                                                                   Figure 11: Estimated ideal points of Chief Justice Burger, the median
                                                                   justice, and Justices Blackmun and Rehnquist, 1969-1985 terms. The
                                                                   vertical axis in both plots is the justice’s estimated ideal point. Higher
                                                                   values are more conservative. The dots are the estimated ideal points.125




                                                                                                                                                            28
                                               Justice Breyer                          Justice Murphy




                                  6




                                                                           6
                                  2




                                                                           2
                                  -2




                                                                           -2
                                  -6




                                                                           -6
                                       1990          Term




                                                            2000                1935         Term




                                                                                                    1945




                                              Justice Stewart                          Justice Thomas




                                  6




                                                                           6
                                  2




                                                                           2
                                  -2




                                                                           -2
                                  -6




                                                                           -6
                                       1955   1965   Term




                                                            1975   1985         1990         Term




                                                                                                    2000




                      Figure 12: Estimated ideal points of justices serving ten or more terms
                      between 1937 and 2005 and who remained relatively stable during their
                      service on the Court. The vertical axis in all plots is the justice’s esti-
                      mated ideal point. Higher values are more conservative. The dots are
                      the estimated ideal points and the vertical error bars in each plot are
                      95% credible intervals. 126


the right, Thomas can declare victory. In the 2004 term Scalia was nearly as close to Rehnquist as
he was to Thomas who clearly anchored the extreme right; and in 2005, Scalia was further from
Thomas than Alito128
    Stewart and Breyer are the remaining justices who fail to exhibit much in the way of preference
change. Juxtaposed against Thomas and Murphy, the two held more centrists ideal points. At the
time of his retirement, Stewart’s pivotal role on the Court moved to the fore, leading to speculation
about the extent to which his eventual replacement, O’Connor, would push the Court to the
right.129 Justice Breyer, while never finding himself in the Court’s center, is hardly an extremist in
the mold of a Murphy or Thomas. During the 1994-2004 terms, a period of stability in the Court’s
membership, Breyer supported litigants alleging an abridgment of their rights or liberties in about
60 percent of the 473 cases; that figure for Stevens, the most liberal justice during those terms, was
 128
     See infra Figure 19.
 129
     Steven R. Weisman, Reagan Nominating Woman, NY Times, July 8, 1981, at 1A. (“White House officials were
hopeful that Judge O’Connor’s appointment could be historic not only because she is a woman but also because her
presence on the Court, as a replacement for Associate Justice Potter Stewart, who was often a swing vote between
ideological camps on the Court, could shift the Court’s balance to the right.”); Stuart Taylor, Jr., Rather an Unknown,
NY Times, July 8, 1981, 13A (“it appears to be far too early to determine whether the ideologically divided Court
will become more conservative or more liberal if Judge O’Connor fills the vacancy created by the retirement of
Justice Potter Stewart, who has been viewed as a moderate leaning to the conservative side of the Court’s delicate
philosophical balance.”); David S. Broder, Doing Justice to the Poor, Wash. Post, June 24, 1981, A21 (“The fact
that the President, who does not see any compelling need for the continuation of the Republican-created program
of legal services for the poor, is the same President who will soon be filling Potter Stewart’s ”swing seat” on the
Supreme Court is something to give you pause.”)
   Worth noting is that Stewart himself rejected the title “swing justice.” When asked at a news conference before
his retirement, “You are regarded as a ‘swing’ Justice, one whose opinions are not easily predictable. Do you think
you should be succeeded by someone like that?” Stewart responded: “I’ve never thought of myself as a swing Justice.
I’ve thought of myself as deciding every case correctly, and I’ve never thought in terms of putting a label on myself.”
Excerpts from Stewart’s Session with Reporters, NY Times, June 20, 1981, at 9.



                                                                          29
over 70.130 A term later, in 2005, Justice Breyer again found himself in the liberal wing, though
its most moderate member.131


                          IV     The Implications of Ideological Change

   The patterns revealed in Figures 5, 7, and 9, however disparate, are conclusive in one im-
portant regard: They cast serious doubt on the commonplace assumption of stable preferences
among Supreme Court justices. At least for justices serving since 1937, ideological drift was not
only possible, it was likely. Of the 26 justices we examined only four did not exhibit significant
fluctuation.
    Certainly the patterns of change differ. While the plurality shifted to the left, a consequential
number moved to right or swung back and forth. Further, change appears to occur at different points
in the justices’ careers. More than a few exhibit what political scientists call, alternatively, the “first-
year,” “freshman,” or “newcomer” effect; that is, an initial period of volatile or uncharacteristic
behavior followed by stability in preferences.132 Earl Warren may fall into this category. After his
first term or so, he moved to the left—and never turned back.
    These patterns deserve consideration, as does the question of what precipitated the observed
changes. In other words, apart from idiosyncratic factors—such as Blackmun’s rift with Burger—
can we identify any underlying, and universal, explanations of change on the Court? We have hinted
at some throughout this article, chiefly the political environment in which the justice operates. Is
it a coincidence that Chief Justice Burger’s most liberal terms came while Jimmy Carter, the
only Democratic President during Burger’s tenure, was in office; and that his most conservative
overlapped with the Reagan years? Likewise, researchers have speculated that both Justices Black
and White may have engaged in “strategic adaptation”: the former moved to the right during
the Nixon presidency; the latter grew more liberal when Kennedy and Johnson were in office and
increasingly conservative during the Nixon and Reagan years.133
    Additional explanations abound,134 and we certainly commend to others the task of exploring
them. For now we focus on the implications of our findings—implications that require only evidence
of change to develop, and not an underlying causal explanation (assuming one exists). Returning
to our earlier discussion, we see two as particularly intriguing: the consequences of change for the
appointment of justices and for doctrinal change. The first implicates the timing of ideological drift
and the second, its importance.
 130
     Computed using Spaeth, supra note 35, with dec type=1, 6, or 7; analu=0; and value ≤ 6.
 131
     See infra Figure 19.
 132
     See, e.g., J. Woodford Howard, Justice Murphy: The Freshman Years, 18 Vand. L. Rev. 473 (1965); Timothy
M. Hagle, “Freshman Effects” for Supreme Court Justices, 37 Am. J. Pol. Sci. 1142 (1993); Terry Bowen & John
M. Scheb, II, Reassessing the “Freshman Effect”: The Voting Bloc Alignment of New Justices on the United States
Supreme Court, 1921-90, 15 Pol. Behavior 1 (1993).
 133
     See Epstein, et al., supra note 120.
 134
     These include context (the justices push and pull each other to the right or left), public opinion (the justices
fluctuate in line with the public’s ideological mood), election returns (the justices follow the election returns), and
the “(Linda) Greenhouse” effect (justice move to the left to win the approval of the New York Times’s Supreme
Court reporter). See, e.g., Ruger, supra note 10; Baum, supra note 112; Ulmer, “The Longitudinal Behavior of Hugo
Lafayette Black,” supra note 44; Epstein, et al., supra note 46.




                                                         30
       A   The Appointments Process and the Timing of Change

    Why the appointment of Supreme Court justices is now, and always has been, a process rife with
political considerations is a question with many answers. But surely one, as we noted earlier, is the
belief among all the relevant actors—the President, senators, interest groups, and the public—that
their choice is particularly weighty. “Because it is nearly impossible to remove a justice, we must
go to lengths to ensure the appointment of the right person, that is, the justice who shares our
ideological commitments, and will for the foreseeable future.” Or so the calculus goes.
    What our results suggest is that predicting the future ideology of any given nominee may be a
risky business, but how risky? That is to say, we know predictions about the long-term ideology
of justices may be highly uncertain in the presence of change. But suppose the relevant political
actors are interested in appointing justices who will reflect their ideological values for, say, a decade.
Is it possible to make accurate forecasts? Succinctly, are the changes we observe in Figures 5, 7,
and 9 more likely to occur later, rather than sooner, in a justice’s tenure?
    As it turns out, the news for political actors is mixed. On the upside, senators and the President
can be reasonably certain that the justice they appoint will behave in line with their expectations—
at least during the justice’s first term in office. Nicely making this point is Figure 13, which plots
the results of comparing the justice’s first- and tenth-term ideal point estimates (see Figures 5-12)
with newspaper editors’ assessments of the justices’ ideology at the time of appointment (see Figure
1).135 The closer a justice is to the line, the better the initial ideological assessment corresponds to
the justice’s first- (left panel) or tenth- (right panel) term revealed preferences.




 135
    To derive Figure 13 we use linear regression to predict the Martin-Quinn estimates of justices in the first and
tenth terms using newspaper editors’ assessments of the justices’ ideology at the time of appointment (see Figure
1). The table below presents the results (standard errors are in parentheses); a visual depiction of this relationship
appears in Figure 13.

                                                           First Term                       Tenth Term
               Intercept                                         0.386                            0.367
                                                               (0.193)                          (0.276)

               Predicted Ideal Point                             1.602                             1.464
                                                               (0.294)                           (0.427)
                                                                n = 28                            n = 26
                                       Residual standard error = 1.021   Residual standard error = 1.402




                                                          31
                                                                              First Year on Court                                                                                                                          Tenth Year on Court
                     Very Conservative




                                                                                                                                                                     Very Conservative
                                                                                                                                                                                                                                                                       Rehnquist q

                                                                                                                     Rehnquist q
                                                                                                                                                                                                                                                               Thomas q

                                                                                                                                                                                                                                                                                  Scalia
                                                                                                                                                                                                                                                                                     q

                                                                                                                       q                                                                             Harlan
                                                                                                                     Thomas                                                                             q




                                                                                                            Blackmun q Burger
                                                                                                                           q



                                                                                                O'Connor         Powell                                                                                                                                                  Burger
                                                                                                                           Scalia
Actual Ideal Point




                                                                                                                                                Actual Ideal Point
                                                                                                   q               q
                                                                                                                             q                                                                                                                            Burton             q

                                                                                                   Kennedy    Roberts q q                                                                                                                                      q
                                                                                                      q              Alito                                                                                                                   O'Connor
                                                                                            q
                                                                                                        q Souter
                                                                                                                                                                                                                  Frankfurter
                                                           q         q
                                                                                         Clark                                                                                            Jackson                             q
                                                                                                                                                                                                                                                   q
                                                                                                                                                                                                                                                                         q
                                                         Harlan Stewart                                                                                                                       q
                                                                                                                                                                                                                                         White                         Powell
                                                                                                           Burton                                                                                             Stewart                       q
                                                                                                                                                                                                                                      Clark q
                                                                                                                                                                                                                                                  q

                                                                                                             q                                                                                                   q                             Kennedy
                                                  q
                                                                                                            Stevens                                                                                           Reed q
                                              Jackson             Warren                 White                   q
                                                                     q
                                                                                            q                                                                                                                                                                                q


                                                                 Ginsburg q                                                                                                                                                                                             Blackmun
                                                                                      Breyer q                                                                                                                                                                     q
                                         Brennan q                                                                                                                                                                                                            Stevens
                                                                                                                                                                                          Brennan
                                         Marshall q                  Reed                                                                                                                     q
                                                                       q

                                                                      Frankfurter                                                                                                                             Warren                         q            q
                                                                              q                                                                                                                                  q

                                                                                                                                                                                                                     q   Douglas          Breyer       Souter
                                                  q   Murphy                                                                                                                                  q

                                                                                                                                                                                           Murphy
                                                                                                                                                                                                        q                 q

                                                                                                                                                                                                     Black        Ginsburg
                     Very Liberal




                                                                                                                                                                     Very Liberal
                                                                                                                                                                                              q

                                                                   Douglas                                                                                                                Marshall
                                                                         q
                                                         Black
                                                           q




                                            Very Liberal                                                                   Very Conservative                                             Very Liberal                                                                        Very Conservative
                                                                                  Predicted Ideal Point                                                                                                                           Predicted Ideal Point



                                                                             Figure 13: Actual (Martin-Quinn) ideal points during a justice’s first and
                                                                             tenth terms plotted against predicted ideal points (based on newspaper
                                                                             editors’ assessments of ideology prior to confirmation). The superim-
                                                                             posed lines are from least squares linear regressions fit to these data
                                                                             (see supra note 135). The closer a point is to the line, the better the
                                                                             prediction. The scale of axes in each plot is identical.136

    If we assume that newspaper editors accurately capture senators’ and the President’s beliefs
about the ideology of their appointees—and there is little reason to think otherwise—then Figure
13 suggests that these elected actors can have some confidence in their beliefs in the very short
term (i.e., the first-year of service). Note how tightly most justices cluster around the first term
regression line—even those justices who later made significant moves to the right or left.137 Justice
Blackmun provides a case in point. Martin and Quinn estimate his ideal point in 1970, his first
term on the Court, at a relatively conservative 1.86. Based on the newspaper editors’ assessments,
we would expect an ideal point of 1.69138 —a trival difference between the actual and predicted
values. In words, Richard Nixon was not wrong, at least not initially, to think that in Blackmun
he was naming a moderately conservative justice to the Court.
    Ten years after appointment, the picture clouds considerably. Underscoring this point is the
righthand panel of Figure 13 in which we can observe that while the aggregate relationship (as
given by the regression line) between editorials at the time of appointment and revealed preferences
remains about the same, the amount of uncertainty about the location of an individual justice (as
 136
     For more detail on the actual (Marti-Quinn) estimated ideal points, see supra notes 85 and 86; for information
on the editors’ assessments, see supra note 37. We generated the predictions via Clarify. See Gary King, Michael
Tomz, & Jason Wittenberg, Making the Most of Statistical Analyses: Improving Interpretation and Presentation, 44
Am. J. Pol. Sci. 341 (2000).
 137
     Further, the most notable departures from the regression line are for justices predicted to be at the extremes of
the ideological spectrum. This is largely due to the fact that the editorial predictions are bounded above and below
with many justices at either the maximal or minimal value, while the Martin-Quinn ideal points are theoretically
unbounded. In other words, much of the variability in the plot is likely an artifact of the scaling of the x and y
variables.
 138
     We estimated this prediction using Clarify, King et al., supra note 136. The 95% confidence interval is [1.05,
2.33].



                                                                                                                                               32
given by the amount of variability around the regression line) increases substantially.139 Which
brings us to the downside for the appointing President and his supporters in the Senate: Even
though the association remains fairly strong, we observe a degradation in the relationship between
the justices initial attitudes and their ideological preferences as soon as ten years out.
    Further proof of the obstacles confronting Presidents in seeking to establish enduring legacies
comes in Figure 14. There we visually depict the probability that the justices were more conservative
(or liberal) in the balance of their terms than in their first. If the solid line in each panel is above
the top dashed line, then the justice is significantly more conservative. If that line is below the
bottom dotted line, then the justice is significantly more liberal. The vertical line represents their
tenth year of service.




 139
     The table depicted in supra note 135 confirms this visual analysis more formally. Note that the slope estimate is
quite similar in both the first term and tenth term regressions. The residual standard error, however, increases from
1.021 in the first term regression to 1.402 in the tenth term regression. In other words, the predicted error around
the regression line increased by 37.3% after ten terms.


                                                         33
                     Justice Blackmun                                 Justice Breyer                          Justice Burger




             1.0




                                                         1.0




                                                                                                 1.0
             0.5




                                                         0.5




                                                                                                 0.5
             0.0




                                                         0.0




                                                                                                 0.0
                   1970          1980        1990              1994      1998     2002                  1970       1976        1982

                           Justice Clark                          Justice Ginsburg                            Justice Kennedy




             1.0




                                                         1.0




                                                                                                 1.0
             0.5




                                                         0.5




                                                                                                 0.5
             0.0




                                                         0.0




                                                                                                 0.0
                    1950         1956      1962                 1994      1998     2002                       1990 1995 2000 2005

                      Justice Marshall                           Justice O'Connor                              Justice Powell
             1.0




                                                         1.0




                                                                                                 1.0
             0.5




                                                         0.5




                                                                                                 0.5
             0.0




                                                         0.0




                                                                                                 0.0
                     1970           1980          1990            1985          1995      2005           1972      1978         1984

                          Justice Souter                          Justice Stevens                             Justice Warren
             1.0




                                                         1.0




                                                                                                 1.0
             0.5




                                                         0.5




                                                                                                 0.5
             0.0




                                                         0.0




                                                                                                 0.0
                   1990      1996          2002                1975     1985     1995     2005           1954      1960         1966




                          Justice Black                               Justice Burton                     Justice Frankfurter
             1.0




                                                         1.0




                                                                                                 1.0
             0.5




                                                         0.5




                                                                                                 0.5
             0.0




                                                         0.0




                                                                                                 0.0




                    1940     1950       1960      1970          1946      1950     1954                  1940          1950      1960

                          Justice Harlan                          Justice Jackson                              Justice Reed
             1.0




                                                         1.0




                                                                                                 1.0
             0.5




                                                         0.5




                                                                                                 0.5
             0.0




                                                         0.0




                                                                                                 0.0




                   1954      1960        1966                   1942      1946     1950                   1940            1950

                          Justice Scalia                          Justice Thomas                               Justice White
             1.0




                                                         1.0




                                                                                                 1.0
             0.5




                                                         0.5




                                                                                                 0.5
             0.0




                                                         0.0




                                                                                                 0.0




                          1990          2000                    1992    1996     2000    2004          1960     1970    1980     1990




Figure 14: The probability that a justice is more liberal or conservative
in subsequent terms than in their first term. The vertical axis denotes
the estimated probability. If the solid line in each panel is above the top
dashed line, then the justice is significantly more conservative. If that
line is below the bottom dotted line, then the justice is significantly more
liberal. The vertical line within each panel represents the tenth term of
service. The top grouping shows justices who were significantly more
liberal by their tenth term relative to their first; the second grouping,
significantly more conservative. We exclude the three justices (Bren-
nan, Murphy, and Stewart) who were neither significantly more liberal
nor conservative by their tenth year compared to their first. We also ex-
clude the two justices who were both more liberal and more conservative
(Douglas and Rehnquist) (for their displays see infra note 140).




                                                                         34
    Relative to their first year, all but three justices (Brennan, Murphy, and Stewart) were signifi-
cantly more liberal or conservative by their tenth year.140 Certainly, in some instances the observed
trends are fairly trivial; for example, compared to his first year, Justice Kennedy was more liberal
by his tenth but the effect dissipates shortly thereafter. Also, to be sure, some Presidents would
not have objected to the drift exhibited by their justices. Scalia is a prime example. We can hardly
imagine his appointing President, Ronald Reagan, arguably the most conservative President of
the 20th century,141 complaining about Scalia’s rather quick, significant, and enduring turn to the
right. Nor do we suspect that Thurgood Marshall’s move to the left would have disturbed Lyndon
Johnson, among the most liberal Presidents of the last five decades.142
    For other justices, however, the change was neither trivial, nor one that the appointing Presi-
dent would have applauded. Chief Justice Warren illustrates both points. When Eisenhower nomi-
nated him was to fulfill a campaign—not exclusively to advance his ideological agenda—neither the
President nor the press deemed the new chief an extreme liberal; indeed, the editors’ initial char-
acterization of Warren was almost identical to their assessment of the moderate Potter Stewart.143
In his first term, Warren lived up to expectations: His ideal point estimate puts him closer to the
center of the Court (Justice Clark) than to the then-extreme liberals Douglas and Black. Note too
how close Warren is to the line in Figure 13 indicating that newspaper editors of the day rather
accurately forecast his first year behavior. But few predicted what would happen next; relative to
his first term, Warren was significantly more liberal in all others.
    Justice O’Connor may provide an even more interesting case, and one not atypical in our data.
Up until she reached about the ten-term mark, the difference between the preferences she revealed
in her first year and all others was inconsequential. Thereafter she made a slow and gradual move
to the left, never to return to the high level of conservatism she exhibited during the years Ronald
Reagan, her appointing President, served. To think about it another way, if Reagan officials believed
O’Connor would retire after a decade of service, their choice was safe: the justice they nominated
and the justice who served were ideologically identical. But, as know, O’Connor remained on the
Court another fifteen years, during which time she moved significantly to the left.
   More generally, if all justices served for ten or fewer terms, preference change would be less of a
concern: It was only by (or close to) the decade mark that we observe behavior significantly different
than the first term for nearly ten justices. The fact of it is, however, that most contemporary justices
remain on the Court far longer. Of the thirty-two justices appointed between 1937 and 2004, only
seven served fewer than ten terms.144 For those in our data set, the length of tenure was, on
 140
    As shown in the plots below, Justice Douglas and Chief Justice Rehnquist, on the other hand, were both more
conservative and more liberal.
                                          Justice Douglas              Justice Rehnquist
                                  1.0




                                                              1.0
                                  0.5




                                                              0.5
                                  0.0




                                                              0.0




                                        1940 1950 1960 1970         1970   1980   1990   2000




 141
     We base this claim on Poole’s NOMINATE Common Space scores. See Keith T. Poole, Estimating a Basic
Space From a Set of Issue Scales, 42 Am. J. Pol. Sci. 954 (1998). The scores are available on Poole’s website, at
http://voteview.com/dwnl.htm.
 142
     See Poole, supra note 141.
 143
     See supra, Figure 1.
 144
     Justices James Byrnes, Wiley B. Rutledge, Fred Vinson, Sherman Minton, Charles Whittaker, Arthur Goldberg,


                                                                             35
average, 21.4 years (with a standard deviation of 7.9). Only five justices served fewer than fifteen
terms—and two of the five remain on the Court.145
    Given the trend toward longer terms,146 the message for Presidents, senators, and interest
groups moves into relief: Those believing that they can entrench their views in the Court for
the decades to come are occasionally mistaken.147 In turn, because these political actors cannot
always accurately predict the future, our results may counsel against ideological appointments—
at the least, ideological appointments to the neglect of other factors but especially a nominee’s
qualifications and his or her ability to advance electoral goals. Let us consider each.
    In previous work, we demonstrated that senators (and perhaps their constituents as well) are
now placing greater weight on a candidate’s ideology and less on their merit than ever before.148
Our results here ought prompt a reevaluation of that balance: While ideology can and does change
with time, background credentials do not. We might even go further and suggest that for a President
seeking to leave a lasting legacy to the nation in the form of justices who will continue to exert
influence on the law well after he leaves office, then a nominee’s professional merit should be a
crucial consideration. To be sure, some 20th century appointees thought to be lacking in the
requisite qualifications went on to be great justices. But many more universally acclaimed as great
justices were also universally perceived as exceedingly well qualified at the time of their nomination:
Oliver Wendell Holmes, Benjamin Cardozo, William Brennan and Antonin Scalia, to name just a
few. While the ideological direction of their doctrinal path may have been hard to predict over the
long term, and not always to the President’s liking, their ability to influence the direction of the
law, based in some part on their intellect, was not.
   Perhaps a bit riskier but nonetheless advisable in light of ideological drift may be a (re)emphasis
on candidates who can advance the President’s or his party’s electoral interests. Of course, this
was the path followed by some past Presidents, including Eisenhower with his nomination of both
Brennan and Warren. Anecdotal and historical evidence, however, suggests that not since the
appointment of Sandra Day O’Connor has a President placed more weight on partisan-electoral
motivations than on other considerations.149 Indeed, in O’Connor’s case, Ronald Reagan was
actually fulfilling a campaign promise to appoint the first female justice150 —a promise his advisors
thought would not only promote Reagan’s candidacy but also advance the future electoral prospects
and Abe Fortas.
 145
     The five are Justices Breyer, Burton, Ginsburg, Jackson, and Murphy.
 146
     See Steven G. Calabresi & James Lindgren, Term Limits for the Supreme Court: Life Tenure Reconsidered, 29
Harv. J. L. & Pub. Pol’y, 770 (2006).
 147
     Jeffrey A. Segal, et al., Buyer Beware? Presidential Success through Supreme Court Appointments, 53 Pol.
Research Q. 557, 557 (2000) make a similar point about the President. They demonstrate that in the short run
Presidents often succeed in appointing ideologically like-minded justices but over time “justices on average appear to
deviate . . . away from the Presidents who appointed them.”
 148
     Epstein, et al, supra note 70.
 149
     Some might argue that George H.W. Bush’s appointment of Clarence Thomas, Clinton’s of Ruth Bader Ginsburg,
and George W. Bush’s of Samuel Alito were all designed to advance partisan interests. But we suspect otherwise,
and in the cases of Thomas and Ginsburg, Yalof’s account, supra note 59, supports our suspicions. In both instances,
the administration considered but rejected candidates who may have been of greater value to their party’s cause.
Because of its recency, no studies yet exist on the calculus behind Alito’s nomination. While we suppose it is possible
that the record will eventually show that partisan concerns (e.g., appeasing the right wing) guided the selection of
Alito, far more likely—given the fit between Alito’s and the President’s preferences—is that ideology was the primary
motivating force.
 150
     On October 14, 1980, Reagan promised that “one of the first Supreme Court vacancies in my administration
will be filled by the most qualified woman I can find, one who meets the high standards I will demand for all my
appointments.” Quoted in Elder Witt, A Different Justice: Reagan and the Supreme Court (1986), 33.



                                                          36
of the Republican party.151
    This is not to say that contemporary Presidents have failed to consider candidates who could
have advanced their party’s electoral interests. They have. According to Yalof’s account,152 in
1994 the Clinton administration shortlisted two candidates designed to enhance the Democrats’
commitment to particular constituencies: Judges Jose Cabranes, who had the support of influential
Hispanic organizations, and Amalya Kearse, a black woman. That the nomination ultimately
went to Stephen Breyer, however, shores up our point: electoral considerations often give way to
others—in Breyer’s case, his political values and his support in the Senate. Given our findings
here of ideological drift, not to mention studies suggesting the importance of symbolic politics
to particular electoral constituencies (think: Reagan’s appointment of the first female justice or
President Bush’s ability to speak Spanish),153 perhaps Presidents ought actually appoint and not
just consider appointing justices primarily for electoral reasons.
    Recommendations about emphasizing or, more precisely, reemphasizing qualifications and par-
tisan interests pertain to Presidents seeking to create lasting legacies. For those political actors
more interested in the short term, our findings suggest a different response: it is possible, even
likely, that during their first few years in office, justices will behave in ways anticipated at the time
of their appointment. This is the primary lesson of Figure 13, documenting the strong relationship
between initial ideological assessments of the justices, as nominees, and revealed preferences during
their first term. Of course, for some Presidents, senators, and interest groups these short-term
payoffs may be sufficient to continue to place emphasis on ideology over credentials—especially if
particular nominees, in particular areas of the law, can work to generate quick doctrinal change.


    B    The Possibility and Importance of Doctrinal Change

    Which brings us to a critical juncture in our analysis. We have now spent many pages docu-
menting ideological movement among the justices. In so doing we have noted the whos, the hows,
and the whens but we have reserved for now perhaps the most fundamental question of all: Does
the ideological drift matter? To what extent can it lead to doctrinal fluctuations among individual
justices and even for the Court as a whole?


         1   Doctrinal Change and Individual Justices

   Turning first to the question of individual-level shifts, their import depends on the direction
and magnitude of the ideological movement. Consider, first, an Antonin Scalia, that is, an extreme
conservative who has grown more extreme with time. Was the move of any doctrinal consequence?
   The answer, as Figure 15 indicates, is probably not much. Here we display Scalia’s term-by-
term ideal point estimates; we also display the “cut point” lines for three cases implicating different
 151
     Worth noting, Reagan’s speech of October 14, 1980 in which he pledged to appoint a woman to the Court, see
supra note 150, was designed to counter accusations that, as Reagan phrased them, he is “somehow opposed to full
and equal opportunities for women in America.” Quoted in Witt, supra note 150, at 33. See also Yalof, supra note
59, at 135-136.
 152
     Yalof, supra note 59, at 204-205.
 153
     For an examination of the importance of symbolic politics, see Stephen P. Nicholson, et al., Ich bin ein Latino!
Sophistication, Symbolism, Heuristics, and Latino Preferences in the 2000 Presidential Election, presented at the
2002 annual meeting of the American Political Science Association, Boston, MA, August 29-September 1.



                                                         37
areas of the law: Lawrence v. Texas,154 Grutter v. Bollinger,155 and Shafer v. South Carolina 156
These lines provide information about the likely behavior of justices above and below it, such that
if a justice’s ideal point is above the line, the probability is greater than .50 that she or he will cast
a conservative vote (i.e., against Lawrence, the Michigan Law School, and Shafer).157 For ideal
points below the line, we predict odds greater than 50-50 that the justice will rule in the liberal
direction (i.e., in favor of Lawrence, the Michigan Law School, and Shafer).
    In the case of Lawrence, we know the Court struck down the sodomy law at issue—an outcome
correctly anticipated by our ideal point estimates: at least five justices fell below the cutpoint
line. We also know that Scalia was not among this group; his estimated ideal point in 2003 was
above the line and, in fact, he dissented in Lawrence. But also note the location of his ideal points
in all previous years. Because they are above the line, we can safely conclude that even at his
most moderate moment—coinciding with the onset of his tenure—Scalia would have likely voted to
uphold the sodomy law. His ideological change, in other words, failed to translate into important
legal change. More generally, in looking at all three cases depicted in Figure 15, in only Shafer
and for only three terms at that would we predict a different response had the case come earlier in
Scalia’s tenure.
    Of course we have not scrutinized the cut points of all cases resolved since the 1986 term when
Scalia joined the Court. But we suspect that additional analyses would only confirm the basic
lesson of Figure 15. Because Scalia was so extreme in his preferences from the start of his service,
 154
      539 U.S. 558 (2003) (striking down sodomy laws).
 155
      539 U.S. 244 (2003) (upholding Michigan Law School’s affirmative action program).
  156
      532 U.S. 36 (2001) (ruling that under the state’s capital sentencing scheme jurors should be informed of a
defendant’s parole ineligibility).
  157
      We derive these cut points using the Martin-Quinn method. Under their approach, the data and modeling
assumptions determine the joint distribution of the ideal points and the cut points. While this joint distribution is
large and complex, it is possible to use the conditional distributions of the ideal points—given the cut points—and
the cut points—given the ideal points—to fit the model, as well as to gain some intuition about how Martin & Quinn
determine the cut points and ideal points.
   To begin, suppose we know the locations of all the cut points. In other words, we know that all justices with an
ideal point to the left of the cut point will be more likely to vote in the liberal direction and all justices to the right
of the cut point will be more likely to vote in the conservative direction. If we observe only one case, then knowledge
of the lone cut point tells us only that some justices (those who voted in the liberal direction on the case) are likely
to be to the left of the cut point and other justices (those who voted in the conservative direction) are likely to be to
the right of the cut point; we cannot infer the location of each justice other than that they are probably somewhere
to the left or right of the cut point. When observe multiple cases, however, and the cut points are treated as known,
more (probabilistic) constraints are applied to the location of the ideal points and tighter estimates of the ideal points
become possible.
   On the other hand, if we treat the ideal points as known we can make inferences about the likely location of the cut
points. To see this, suppose we observe the following sequence of votes (ordered from left to right), where L denotes
a liberal vote and C a conservative vote:

                                                  LLLCCCCCC

  From this sequence, we would infer that the most likely place for the cut point would be somewhere between the
third and the fourth justice. (The exact location is determined by the particular modeling assumptions employed
but it is qualitatively similar across a range of reasonable alternative assumptions.) Cases with equivalent observed
voting patterns will have the same estimated cut points.
  By alternately conditioning on the cut points to infer the conditional distribution of the ideal points and conditioning
on the ideal points to infer the conditional distribution of the cut points Martin & Quinn are able to take a sample
that is approximately from the joint distribution of the ideal points and cut points given the observed votes on the
merits.
 158
     For more detail on cut points, see supra note 157.



                                                            38
                                                           3.5
                                                                                                               q q
                                                                                                         q q         q
                                                                                                     q




                                                           3.0
                                                                                                 q                       q
                                                                                             q                               q




                                                           2.5
                             Estimated Ideal Point Scale
                                                                                     q q q                                       q q




                                                           2.0
                                                                                 q
                                                                             q
                                                                                                 Shafer Cutpoint
                                                                         q
                                                                     q


                                                           1.5
                                                                 q

                                                           1.0

                                                                                                 Lawrence Cutpoint
                                                           0.5




                                                                                                 Grutter Cutpoint
                                                           0.0




                                                                             1990            1995               2000              2005

                                                                                                 Term



                      Figure 15: Time series plot of Justice Scalia’s estimated ideal points,
                      1986-2005 terms. The horizontal lines are the cut points for Grutter
                      v. Bollinger, Texas v. Lawrence, and Shafer v. South Carolina such
                      that points above the line indicate a probability of greater than .50 of
                      voting conservatively; those below the line indicate a greater than .50
                      probability of voting in the liberal direction (as the Court did in each of
                      the depicted cases).158


his turn to the right indicates only a marginal change in his jurisprudence. The same likely holds
for Justices Brennan and Marshall—other rather extremists (though liberals) who only grew more
extreme over time.
    For the balance of our justices, however, ideological fluctuations may well have precipitated
doctrinal change of some consequence. Again, Harry Blackmun provides the most obvious case in
point—with Figure 16 providing but one example. There we display his ideal points along with
the estimated cutpoints for two landmark death penalty cases, Furman v. Georgia 159 and Gregg v.
Georgia.160 When the Court decided in Furman to strike down all existing death penalty statutes,
Blackmun dissented. Given that his revealed preferences for the 1971 term were north of the Furman
cutpoint line, the dissent was not a surprise. And neither, for that matter, was his concurrence
in Gregg supporting the Court’s decision to uphold newly fashioned capital punishment laws. His
ideal point remained above the Gregg cutpoint line.
    Seen in this way, Blackmun provides an example of how Presidents, allied senators and support-
ing organized interests can gain short-term policy benefits from appointing ideologically compatible
nominees. When Nixon nominated Blackmun to the Court, the President believed his new justice
was committed to a law-and-order stance, and newspaper editors of the day agreed. In Furman
and Gregg, Blackmun did not disappoint. Note, though, that by 1976 Blackmun’s ideological shift
began to seep into his death penalty jurisprudence. Had the Court decided Furman in 1976, the
 159
     408 U.S. 238 (1972).
 160
     428 U.S. 15 (1976).
 161
     For more detail on cut points, see supra note 157.



                                                                                             39
                                                              2
                                                                    qq

                                                                          q
                                                                              q
                                                                                  q




                                                              1
                                                                                      q
                                                                                                  Furman Cutpoint




                                Estimated Ideal Point Scale
                                                                                          q

                                                                                              q
                                                                                                  q




                                                              0
                                                                                                      q              q
                                                                                                          q              q
                                                                                                              qq                 Gregg Cutpoint

                                                                                                                             q
                                                                                                                                       q
                                                                                                                                 qqq
                                                              −1


                                                                                                                                           q
                                                                                                                                               q
                                                                                                                                                   q
                                                                                                                                                       q

                                                                   1970           1975                1980               1985            1990

                                                                                                              Term



                         Figure 16: Time series plot of Justice Blackmun’s estimated ideal points,
                         1970-1994 terms. The horizontal lines are the cut points for Furman v.
                         Georgia and Gregg v. Georgia such that points above the line indicate
                         a probability of greater than .50 of voting against the defendant (as the
                         Court did in Gregg); those below the line indicate a greater than .50
                         probability of voting for the defendant (as the Court did in Furman).161


probability of Blackmun upholding Georgia’s death penalty law would have fallen below .50; and
had it decided Gregg after 1985, Blackmun would likely have voted to strike the new statutes.
    Because Blackmun moved from a rather extreme conservative to a rather extreme liberal, the
effect of his ideological turnabout on doctrine is especially noticeable. But the trend need not be
as dramatic as Blackmun’s for it to manifest in legal change. Chief Justice Rehnquist’s turn to the
center, as we show in Figure 17, provides an interesting example. Here we display cutpoints for
Lawrence, Grutter, and Shafer, as well as for Wiggins v. Smith,162 in which the Court held that
the defendant’s attorney had had failed to provide effective counsel during the sentencing phase
of his capital case. Observe that in neither Grutter nor Lawrence did Rehnquist’s leftward trend
translate into doctrinal change: Odds are that at no point in his career would he have voted to
strike the sodomy law at issue in Lawrence or uphold the affirmative action program in Grutter.
And, in fact, he dissented in both cases. The two capital cases present a different picture. Had
either been before the Court prior to the early 1990s, we predict that Rehnquist would have ruled
for the state in both. But thereafter he had moved sufficiently to the left that the odds shifted in
favor of the defendant. And, in fact, in both Shafer and Wiggins Rehnquist cast votes against the
government.
 162
       539 U.S. 510 (2003).
 163
       For more detail on cut points, see supra note 157.




                                                                                                          40
                                                                     qq
                                                                   qq qqqq




                                                           4
                                                                  q       qq
                                                                            q
                                                                 q           q
                                                                              q
                                                                               q




                             Estimated Ideal Point Scale

                                                           3
                                                                                                 q
                                                                                                  q
                                                                                                   q
                                                                                                    q



                                                           2
                                                                                                     qq
                                                                       Shafer Cutpoint
                                                                                                       qqq qqq
                                                                       Wiggins Cutpoint                   qq     q
                                                                                                              q q
                                                                                                               q
                                                           1


                                                                       Lawrence Cutpoint


                                                                       Grutter Cutpoint
                                                           0




                                                               1970   1975    1980        1985     1990   1995   2000   2005

                                                                                             Term



                      Figure 17: Time series plot of Justice Rehnquist’s estimated ideal points,
                      1970-2004 terms. The horizontal lines are the cut points for Grutter v.
                      Bollinger, Texas v. Lawrence, Wiggins v. Smith, and Shafer v. South
                      Carolina such that points above the line indicate a probability of greater
                      than .50 of voting conservatively; those below the line indicate a greater
                      than .50 probability of voting in the liberal direction (as the Court did
                      in each of the depicted cases).163


         2   Doctrinal Change and the Court’s Center

    Rehnquist’s shift is interesting if only because it is so unexpected, but its impact on the es-
tablishment of precedent is far from clear. Because both Shafer and Wiggins were decided by 7-2
majorities, the Chief’s vote was likely not necessary for the creation of precedent. More gener-
ally, because Rehnquist never served as the Court’s median or swing justice, even as he grew less
extreme, his shift was less than consequential.
    Not so of the more centrists justices. Take Sandra Day O’Connor. As we already noted, over
the last decade or so, O’Connor trended to the left; less obvious from our analyses, though widely
acknowledged, is that O’Connor’s was the likely median (or swing) justice for over a third of her
service on the Court (nine of twenty-five terms). Even more impressively, she managed to hold that
crucial position for an extraordinary seven consecutive terms.164
    With little doubt these two phenomena—O’Connor’s move to the left and her role as the swing
justice—coalesced to produce noticeable and consequential doctrine during the latter Rehnquist
Court years. Figure 18 makes the point for two cases, Lawrence and Grutter. Here we show ideal
point estimates for O’Connor and for the median Justice over the last two decades. (Solid black
circles indicate terms when O’Connor was the median.) We also show the cutpoints for Grutter
and Lawrence.
 164
    See Martin, et al., supra note 88.
 165
    For more detail on cut points, see supra note 157. Note that 2005 has a black circle indicating that O’Connor was
the median, as well as a triangle indicating that Kennedy moved into the median position when O’Connor departed.



                                                                                            41
                                                                                                                                                 O'Connor
                                                                        q                                                                        Median
                                                                            q




                                                              1.5
                                                                    q                       q
                                                                                q               q
                                                                                                    q
                                                                                        q
                                                                                    q




                                Estimated Ideal Point Scale

                                                              1.0
                                                                                                        q
                                                                                                                    q               q
                                                                                                                                q        Lawrence Cutpoint
                                                                                                                                        qq
                                                                                                                q           q
                                                                                                                        q
                                                              0.5                                           q
                                                                                                                                             q

                                                                                                                        Grutter Cutpoint         q
                                                                                                                                                     q
                                                                                                                                                         q
                                                                                                                                                             q
                                                              0.0




                                                                                                                                                                 q
                                                                                1985                1990                1995              2000               2005

                                                                                                                Term



                         Figure 18: Time series plot of Justice O’Connor’s and the median’s
                         estimated ideal point, 1981-2005 terms. The solid black circles indicate
                         that Justice O’Connor is most likely the median Justice. The horizontal
                         lines are the cut points for Grutter v. Bollinger and Lawrence v. Texas
                         such that points above the line indicate a probability of greater than
                         .50 of voting to strike down the program (Grutter ) and uphold the law
                         (Lawrence); those below the line indicate a greater than .50 probability
                         of voting to uphold the program in Grutter (as the Court did in the 2002
                         term) and strike down the law in Lawrence (again as the Court did in
                         the 2002 term).165


    Certainly Figure 18 should once again dispel any doubt that fluctuation in preferences, and
significant fluctuation at that, is possible. But more that it illustrates, to varying degrees, the po-
tential importance of shifts when it is the median justice who is shifting. Beginning with Lawrence,
notice that in the early part of the natural court period that began in 1994 (and did not end until
2005), O’Connor’s ideal point estimate is close to the cutpoint line, though occasionally above it
(indicating a greater than 50 percent likelihood of upholding the sodomy law). In other words, had
Lawrence come to the Court in, say, the 1987 term, the odds are that O’Connor would have voted
with the dissenters (note that the median in 1997 is below the cutpoint line). By the 1999 term
her gradual turn to the left, coinciding with her capture of the median position, increased the odds
considerably of the Court striking the sodomy law—a step it eventually took.
    O’Connor’s role in the Grutter litigation was even more crucial.166 In the 1994 term (when
Kennedy was the Court’s most likely median) the probability of the Court supporting the the law
school’s affirmative action program was just 0.32. In other words, neither the Court nor O’Connor
would likely have voted in Michigan’s favor had Grutter come in 1994 or even as late as 2000. But
O’Connors turn to the left and into the median position, seven terms later, in 2001, proved decisive
to the outcome of the case.
    It was O’Connor’s growing liberalism, coupled with her role as swing justice, that provides, at
the least, a partial explanation for the decisions in Lawrence but especially Grutter. Which brings
 166
       We make a similar point in Martin, et al., supra note 88.


                                                                                                            42
us to a crucial point. Of the 16 justices in our data set who exhibited significant change and no
longer remain on the Court, all but five—the three chief justices (Warren, Burger, and Rehnquist)
and associates William O. Douglas and Robert Jackson—served as the median justice.167 If as the
moved they took the Court with them, then they provide the clearest evidence of ideological change
translating into doctrinal change.
    The implications of these results are several, but surely one is that lawyers make assumptions
about the (im)possibility of legal change during periods of membership stability to their own dis-
service. Figure 18, we believe, drives home this point with force. The period between the 1994 and
2004 terms may have been one of the longest natural courts in American history but the revealed
preferences of individual justices, including the median, changed rather markedly. Not surprisingly,
important doctrinal shifts came in their wake.


       V   Preference Change and Doctrinal Development on the Roberts Court

    With the arrival of Chief Justice Roberts and Justice Alito, will more shifts follow—even in the
absence of further membership change? We suspect so. In the first place, while both new justices,
not unexpectedly, have emerged as conservative figures (see Figure 19), ideological drift is hardly
impossible; actually, based on our results here, it is likely. But whether one or both will exhibit
abrupt change, as did Earl Warren or a more gradual trend, as did O’Connor, or even in what
direction they will move, we cannot say in the absence of a theory of preference change. All we can
suggest for now is that neither is likely to stay his current ideological course for a decade or longer.
    Second, and here we can be more concrete, with the arrival of Roberts and Alito, the Court’s
center has shifted slightly to the right—from O’Connor to Kennedy (see Figure 19). With that shift,
doctrinal change is likely to follow regardless of whether Kennedy adheres to his current doctrinal
posture or drifts further to the left.
    To see why, first consider a scenario under which no new justices join the Court. Also assume
(in contrast to our overall findings here) that the justices’ current ideal point estimates remain
relatively stable. Under these assumptions, and given the configuration of preferences displayed in
Figure 19, Justice Kennedy will hold the swing position for the foreseeable future—meaning that
doctrinal development rests largely on his shoulders.168 More specifically, for areas of the law in
which he and O’Connor were below (or above) various cut points, we would predict minimal legal
change. Along these lines is Lawrence. The odds in 2003 were far greater than 50-50 that both
would vote to strike down the law; and, now in 2006, as we show in Figure 20, Kennedy remains
well south of the Lawrence cut point line.
    Legal change here is thus highly unlikely but not so for affirmative action. As Figure 20 also
displays, at no point in his career did Kennedy’s revealed preferences fall below the Grutter line.
Indeed, today the odds are only about 34 percent that he would vote to uphold a Grutter -like
program. Given his current role as the median justice, and again assuming no preference change
 167
     See Martin, et al., supra note 88.
 168
     Some commentators have gone so far as to deem the 2005 term the onset of the Kennedy—and not Roberts—
Court era. See, e.g., The Fragile Kennedy Court, N.Y. Times, July 7, 2006, 16A (“The Supreme Court has nominally
been the Roberts Court since last fall, when John Roberts arrived as chief justice. But as a practical matter, the
recently completed term marked the start of the Kennedy Court.”).
 169
     For more detail on cut points, see supra note 157.




                                                       43
                               1.0

                               0.8




                  Separation
                               0.6

                               0.4

                               0.2

                               0.0   Stevens




                                                          Breyer




                                                                                Kennedy
                                               Ginsburg
                                                 Souter




                                                                                                            Scalia




                                                                                                                     Thomas
                                                                     O'Connor




                                                                                          Roberts / Alito
                                                          Estimated Ideal Point Scale



                         Figure 19: Justices serving on the 2006 term Court, and Justice
                         O’Connor, arrayed according to their 2005 term ideal point estimates.
                         The vertical lines are at the mid points between contiguous pairs of jus-
                         tices and the height of these vertical lines gives the separation between
                         these two justices. Here we define separation as 2 × [Pr(justice i + 1
                         is to the right of justice i) - 0.5]. Thus, a separation of 0 indicates the
                         justices occupy the same position and a separation of 1 indicates there
                         is no overlap whatsoever.


among the existing justices, this will come as disturbing news for supporters of affirmative action
in education and, of course, a promising development for opponents.
    And yet, the evidence of widespread ideological drift we have offered here suggests that this
status quo scenario, while not impossible, is unlikely. Far more plausible is a scenario in which at
least one justice exhibits ideological fluctuation. New Court members are always prime suspects.
As we have seen, it is difficult to make inferences about their long-term patterns based on their
first-year preferences. But even setting aside Alito and Roberts, doctrinal change (or, in some
instances, surprising stability) is possible if Kennedy continues to drift to the left.
    Or, more precisely, if Kennedy renews his leftward drift. As we noted earlier, while Kennedy is
significantly more liberal now than in the late 1980s, his ideal point has remained relatively flat over
the last decade or so. Hence, whether Kennedy has reached the zenith of liberalism or will come to
resemble an an O’Connor—a justice who exhibited a gradual, though highly consequential, shift—
we cannot say. What we can claim is that at least in some areas of the law, change on Kennedy’s
part must be rather dramatic for it to exert influence on the course of doctrine. To return to
Lawrence, the odds today are so slim of Kennedy voting in favor of at least certain kinds of laws
discriminating against gays that only a seismic shift (and to the right, at that) would reverse them.
   In other legal areas, however, even small leftward movement on Kennedy’s part may be no-


                                                                    44
                                                           1.4
                                                                     q
                                                                         q




                                                           1.2
                                                                 q            q




                                                           1.0
                             Estimated Ideal Point Scale
                                                                                                                       q          q
                                                                                        q                                   q




                                                           0.8
                                                                                            Lawrence Cutpoint                         q
                                                                                    q       q                      q                      q
                                                                                                q          q


                                                           0.6
                                                                                                               q
                                                                                                     q
                                                                                                                                              q
                                                                                                                                                   q
                                                           0.4


                                                                              Grutter Cutpoint
                                                           0.2




                                                                              McCreary Cutpoint
                                                           0.0




                                                                             1990                   1995                   2000                   2005

                                                                                                         Term



                      Figure 20: Time series plot of Justice Kennedy’s estimated ideal point,
                      1986-2005 terms. The horizontal lines are the cutpoints for Grutter v.
                      Bollinger, Lawrence v. Texas, and McCreary County v. American Civil
                      Liberties Union such that points above the line indicate a probability of
                      greater than .50 of voting to strike down the program (Grutter ), uphold
                      the law (Lawrence) or allow the display (McCreary); those below the line
                      indicate a greater than .50 probability of voting to uphold the program
                      in Grutter (as the Court did in the 2002 term), strike down the law
                      in Lawrence (again as the Court did in the 2002 term), or disallow the
                      display (the step taken by the Court in McCreary).169


ticeable. For O’Connor one of those areas was affirmative action (see Figure 18), and we might
say the same of Kennedy. While our current estimates place him above the Grutter cut point
line, the distance between it and his revealed preferences is narrowing in much the same way as
did O’Connor’s over time. In other words, should Kennedy trend left, concerns about the demise
of affirmative action may dissipate. Even assuming that the two new justices prefer to overturn
Grutter, the Court’s new center could preserve it, thereby generating doctrinal stability rather than
change.
    Surprising stability also might result in yet another contentious area, religious establishment,
particularly the display of religious symbols. In the 2005 case of McCreary County v. American
Civil Liberties Union,170 a five-to-four Court held that the display of the Ten Commandments
in county courthouses violated the establishment clause. Because Justice O’Connor was in the
majority, and Justice Kennedy, in dissent,171 some commentators have suggested this is an area
ripe for legal change.172
 170
     545 U.S. 844 (2005).
 171
     Interestingly, commentators do not regard Justice O’Connor as the swing vote in the case; they, instead, point
to Justice Breyer. The reason is that on the same day the Court handed down McCreary, it also decided Van Orden
v. Perry, 545 U.S. 677 (2005), in which it upheld the display of the Ten Commandments on the grounds of the Texas
state capitol. Breyer was the only justice in the majority in both. See, e.g., Paul Gewitz, The Pragmatic Passion
of Stephen Breyer 115 Yale L.J. 1675,1693 (2006) (“The Ten Commandments cases are especially noteworthy
because Breyer ended up being the pivotal Justice in each case, providing the decisive fifth vote.”).
 172
     Erwin Chemerinsky, The End of an Era, 8 Green Bag 2d 345, 352 (2005) (“With four Justices—Rehnquist,


                                                                                                     45
    Based on our analysis, they are not wrong. Both new justices, and Kennedy himself are above
the McCreary cut point line (see Figure 20), indicating that, in all likelihood, the McCreary dis-
senters would prevail if the case were reheard today. On the other hand, should Justice Kennedy
renew his drift to the left, McCreary and its progeny may be in less jeopardy than some suspect.
As we show in Figure 20 Kennedy’s revealed preferences are creeping promisingly or perilously,
depending on one’s perspective, toward the cut point line. This is not to suggest that doctrine
governing this area will remain unchanged. In fact, if Kennedy has his way, the Court will revisit
the standards it uses to resolve these disputes.173 It is rather to suggest that, even with a new test,
the outcomes may be less dramatically different than some predict.


                                                  VI     Conclusion

    Throughout his tenure Justice Harry A. Blackmun repeatedly told interviewers that it was the
Court, and not he, who had changed.174 In 2005 President Bush asked Americans to trust him, that
Harriet Miers would not change once appointed to the Court.175 And in the same year Professor
David Strauss told us to disbelieve claims that John G. Roberts, Jr. would moderate upon his
elevation to the high Court.176
    Does the evidence bear out their claims? Yes and no. On the one hand, our results suggest that
a close relationship exists between our expectations about a nominee’s ideology and the ideology
they reveal during the first few term in office. Data from newspaper editorials suggested that Earl
Warren and David Souter would be moderate-to-conservative in their ideological outlook, and they
Scalia, Kennedy, and Thomas—eager to overrule the Lemon test and allow a much greater presence of religion in
government, this is an area where Justice O’Connor’s successor could have an immediate and dramatic effect on the
law.”); Marci A. Hamilton, The Establishment Clause During the 2004 Term: Big Cases, Little Movement, 2004-05
Cato Sup. Ct. Rev. 159, 184 (2004/2005) (noting that the differences between the majority and dissenters in
McCreary are “stark,” and that “and the next justice of the Supreme Court, who replaces Justice O’Connor, will have
the power to shift the doctrine either way.”); Marcia S. Alembik, The Future of the Lemon Test: A Sweeter Alternative
for Establishment Clause Analysis, 40 Ga. L. Rev. 1171, 1207 (2006) (“Three justices, Justices Scalia, Kennedy,
and Thomas have already indicated that they think the Lemon test should be overruled, and the addition of two new
Justices, both sharing this view, could cause the Lemon test to turn sour.”); Christopher B. Harwood, Evaluating the
Supreme Court’s Establishment Clause Jurisprudence in the Wake of Van Orden v. Perry and McCreary County v.
ACLU, 71 Mo. L. Rev. 317, 348 (2006) (“The appointment of Chief Justice Roberts and Justice Alito to fill the
vacancies left by Chief Justice Rehnquist and Justice O’Connor likely will alter the Court’s Establishment Clause
jurisprudence and produce decisions that conform to the teachings of the accommodation approach. Last term, the
neutrality approach enjoyed majority support by the slimmest of margins, and one of the supporters of that approach,
Justice O’Connor, has since left the Court.”)
  173
      Justice Kennedy has advocated a coercion test to resolve religious establishment disputes. See Lee v. Weisman,
505 U.S. 577, 587 (1992) (“It is beyond dispute that, at a minimum, the Constitution guarantees that government
may not coerce anyone to support or participate in religion or its exercise”); County of Allegheny v. ACLU, 492
U.S. 573, 660 (1989) (“Our cases disclose two limiting principles: government may not coerce anyone to support or
participate in any religion or its exercise; and it may not, in the guise of avoiding hostility or callous indifference,
give direct benefits to religion in such a degree that it in fact ’establishes a [state] religion or religious faith, or tends
to do so.’”). Under this approach, he has voted to uphold (Allegheny) and strike down (Lee) practices challenged
under the establishment clause. See, e.g., Cynthia V. Ward, Coercion and Choice Under the Establishment Clause,
39 U.C. Davis L. Rev. 1621 (2006); William Van Alstyne, Nine Judges, and Five Versions of One Amendment, 14
Wm. & Mary Bill of Rts. J. 17, 29 (2005) (“Justice Kennedy, while generally more disposed to the generic view
common to Rehnquist, Scalia, and Justice Thomas, is nevertheless quite at odds with them when he finds evidence
that government has brought some degree of ‘coercion’ to bear in its various religious preferments.”).
  174
      See Jenkins, supra note 11.
  175
      See supra note 24.
  176
      Strauss, supra note 7.



                                                             46
were not wrong. In their freshman year, both voted in accord with that label. So when Professor
Strauss tells us to ignore predictions about the possibility that Roberts will moderate he may well
be right in the short term.
    On the other hand, our results indicate that Professor Strauss may be mistaken in the longer
term. The ideological boxes into which Presidents, senators, and the public place justices at the
time of their nomination are not so tightly sealed. Drift to the right or, more often, the left is the
rule, not the exception. In some instances, the movement may be relatively inconsequential but in
others substantial doctrinal change may result. Grutter provides a powerful example, and it is by
no means the only one.
    And because, we suspect, it is by no means the last one, Presidents would well serve themselves
by considering the possibility of drift when making appointments to the Court. Given the potential
tradeoff between long-term ideological control and shorter-term electoral gain, our evidence speaks
to placing comparatively greater emphasis on the latter.177 Certainly, the two are not always in
conflict. With the nomination of Samuel Alito, President Bush appointed a justice who may or
may not remain right-of-center well into the 2010s. Either way, though, the nomination managed
to placate his conservative base. Likewise, even if an Emilio M. Garza or an Edward C. Prado
were to drift from their conservative roots, the President would still have captured the benefits
of nominating the first Latino to the Court—benefits that could have multiplied if the Democrats
chose to fight the nomination. Scholars can label William J. Brennan or Sandra Day O’Connor
failures from the point of view of their appointing President, but both served crucial partisan
interests. The real failures, from the President’s perspective, are those, like Harry Blackmun, who
in the long run drift from their initial preferences without ever having provided electoral benefits.




 177
   Alternatively, given the prominence of leftward drift among recent justices, Presidents ought consider nominees
more conservative than their ideal points.


                                                       47